diff --git a/guides/ai-image-api-rate-limit-handling.md b/guides/ai-image-api-rate-limit-handling.md
new file mode 100644
index 0000000..43da95b
--- /dev/null
+++ b/guides/ai-image-api-rate-limit-handling.md
@@ -0,0 +1,80 @@
+# AI Image API Rate Limit Handling for Batch Workloads
+
+Control image API concurrency with rate-limit headers, bounded queues, jittered backoff, and per-model throughput records.
+
+> Test note: model behavior and pricing change. Record the model ID, parameters, date, latency, retries, and accepted output count for every run.
+
+## Why this guide exists
+
+Batch systems should respond to provider limits without dropping job provenance or creating retry storms, so this guide separates admission control, worker concurrency, backoff, and accepted-output accounting across slow image requests.
+
+## Starter fixtures
+
+### Minimal coffee packaging
+
+**Goal:** Packaging concept
+**Aspect ratio:** 3:4
+
+```text
+Premium whole-bean coffee pouch standing on dark walnut, uncoated cream paper with a small black label area, morning side light, restrained props, realistic folds, front panel fully visible, no invented text.
+```
+
+### Smart speaker studio
+
+**Goal:** Consumer electronics hero
+**Aspect ratio:** 1:1
+
+```text
+Compact smart speaker in graphite fabric on a deep charcoal pedestal, subtle violet rim light, precise mesh texture, controlled reflections, high-end consumer electronics launch photography, centered composition.
+```
+
+### Summer drink social ad
+
+**Goal:** Mobile social creative
+**Aspect ratio:** 4:5
+
+```text
+Bright summer beverage can on crushed ice, coral and turquoise color blocking, hard flash photography, realistic condensation, product occupies the lower two-thirds, clean space above for campaign copy, no text.
+```
+
+## Measured evidence
+
+- [GPT Image vs Flux vs Gemini: Product Photography Test](../comparisons/gpt-image-vs-flux-vs-gemini-product-photography.md) — GPT Image 2 delivered the strongest editorial staging, Flux followed the literal window-light direction most closely, and Gemini produced the cleanest but least distinctive pack shot.
+- [AI Billboard Image Generation Comparison](../comparisons/ai-billboard-image-generation-comparison.md) — GPT most closely matched the minimal no-lettering direction. Flux invented label marks, while Gemini drifted toward a literal outdoor billboard scene.
+
+## Python quickstart
+
+```python
+import os
+from openai import OpenAI
+
+client = OpenAI(
+ api_key=os.environ["APIMART_API_KEY"],
+ base_url="https://api.apimart.ai/v1",
+)
+
+response = client.images.generate(
+ model="gpt-image-1-official",
+ prompt="YOUR_PROMPT",
+ size="1024x1024",
+ n=1,
+)
+```
+
+## Run through one API gateway
+
+APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
+
+[Create an APIMart account and run this workflow](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-register-47639944) | [API documentation](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-docs-en-quickstart-e3cffa72)
+
+## Review checklist
+
+- Did the output follow the requested composition and protected attributes?
+- Were the model identifier, parameters, and test date recorded?
+- How many retries were required for one accepted output?
+- Does the output contain invented text, marks, or unwanted objects?
+- Can another contributor reproduce the request from the stored fixture?
+
+## Disclosure
+
+This open-source benchmark is maintained by APIMart Labs. Links to APIMart are promotional and directly related to running the documented API examples.
diff --git a/guides/async-image-generation-api-polling.md b/guides/async-image-generation-api-polling.md
new file mode 100644
index 0000000..3c6a6a4
--- /dev/null
+++ b/guides/async-image-generation-api-polling.md
@@ -0,0 +1,80 @@
+# Async Image Generation API Polling with Bounded Backoff
+
+Poll asynchronous image-generation jobs with bounded backoff, terminal-state handling, timeouts, and durable request metadata.
+
+> Test note: model behavior and pricing change. Record the model ID, parameters, date, latency, retries, and accepted output count for every run.
+
+## Why this guide exists
+
+Long-running image jobs need a polling loop that distinguishes queued, running, completed, failed, and expired states while limiting unnecessary requests and preserving enough metadata to resume safely after a worker restart.
+
+## Starter fixtures
+
+### Ceramic skincare hero
+
+**Goal:** Clean premium product hero
+**Aspect ratio:** 1:1
+
+```text
+Editorial product photograph of a matte ceramic skincare bottle on pale limestone, soft north-window light, delicate contact shadow, warm neutral palette, 85mm lens compression, no text, no logo.
+```
+
+### Running shoe impact
+
+**Goal:** Dynamic footwear campaign
+**Aspect ratio:** 4:5
+
+```text
+Technical running shoe frozen above a red clay track at impact, fine dust particles suspended in air, directional hard sunlight, visible outsole detail, energetic diagonal composition, commercial sports photography.
+```
+
+### Coastal house at dusk
+
+**Goal:** Exterior visualization
+**Aspect ratio:** 16:9
+
+```text
+Low modern coastal house at blue hour, weathered cedar and board-formed concrete, dune grasses moving in sea wind, warm interior light, realistic architectural photography, human-scale lens.
+```
+
+## Measured evidence
+
+- [AI Running Shoe Ad Model Comparison](../comparisons/ai-running-shoe-ad-model-comparison.md) — GPT created the most polished commercial crop, Flux showed the strongest outsole and track realism, while Gemini removed the athlete and emphasized a floating product shot.
+- [AI Architecture Rendering Model Comparison](../comparisons/ai-architecture-rendering-model-comparison.md) — All three produced convincing architecture. Gemini completed in 32 seconds, while GPT delivered richer detail but required 159 seconds for this fixture.
+
+## Python quickstart
+
+```python
+import os
+from openai import OpenAI
+
+client = OpenAI(
+ api_key=os.environ["APIMART_API_KEY"],
+ base_url="https://api.apimart.ai/v1",
+)
+
+response = client.images.generate(
+ model="gpt-image-1-official",
+ prompt="YOUR_PROMPT",
+ size="1024x1024",
+ n=1,
+)
+```
+
+## Run through one API gateway
+
+APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
+
+[Create an APIMart account and run this workflow](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-register-47639944) | [API documentation](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-docs-en-quickstart-e3cffa72)
+
+## Review checklist
+
+- Did the output follow the requested composition and protected attributes?
+- Were the model identifier, parameters, and test date recorded?
+- How many retries were required for one accepted output?
+- Does the output contain invented text, marks, or unwanted objects?
+- Can another contributor reproduce the request from the stored fixture?
+
+## Disclosure
+
+This open-source benchmark is maintained by APIMart Labs. Links to APIMart are promotional and directly related to running the documented API examples.
diff --git a/guides/image-generation-api-webhook-processing.md b/guides/image-generation-api-webhook-processing.md
new file mode 100644
index 0000000..1ca1237
--- /dev/null
+++ b/guides/image-generation-api-webhook-processing.md
@@ -0,0 +1,80 @@
+# Image Generation API Webhook Processing and Verification
+
+Process image-job webhooks with signature verification, idempotent updates, asset downloads, and replay-safe event records.
+
+> Test note: model behavior and pricing change. Record the model ID, parameters, date, latency, retries, and accepted output count for every run.
+
+## Why this guide exists
+
+Webhook delivery can be duplicated, delayed, or reordered, so the receiver should authenticate each event, preserve its provider identifier, update one durable job record, and download expiring assets only after a verified completion state.
+
+## Starter fixtures
+
+### Ceramic skincare hero
+
+**Goal:** Clean premium product hero
+**Aspect ratio:** 1:1
+
+```text
+Editorial product photograph of a matte ceramic skincare bottle on pale limestone, soft north-window light, delicate contact shadow, warm neutral palette, 85mm lens compression, no text, no logo.
+```
+
+### Forest ranger character sheet
+
+**Goal:** Identity consistency fixture
+**Aspect ratio:** 3:2
+
+```text
+Character sheet for Mara, a forest ranger with cropped auburn hair, olive field jacket, brass compass and worn canvas backpack; front, profile and three-quarter views; neutral studio background; consistent facial features.
+```
+
+### Climate intelligence dashboard
+
+**Goal:** Dense editorial interface
+**Aspect ratio:** 16:10
+
+```text
+Editorial analytics dashboard for a climate intelligence product, dense but legible information hierarchy, off-white canvas, dark graphite type, acid green data accents, modular card grid, desktop viewport.
+```
+
+## Measured evidence
+
+- [Consistent Character AI Model Comparison](../comparisons/consistent-character-ai-model-comparison.md) — All three maintained the ranger identity across views. GPT added the richest close-up and prop detail; Flux and Gemini produced cleaner, simpler turnaround sheets.
+- [AI Dashboard UI Generation Comparison](../comparisons/ai-dashboard-ui-generation-comparison.md) — GPT produced the most legible information hierarchy. Gemini offered the strongest cost-latency compromise, while Flux contained more pseudo-text.
+
+## Python quickstart
+
+```python
+import os
+from openai import OpenAI
+
+client = OpenAI(
+ api_key=os.environ["APIMART_API_KEY"],
+ base_url="https://api.apimart.ai/v1",
+)
+
+response = client.images.generate(
+ model="gpt-image-1-official",
+ prompt="YOUR_PROMPT",
+ size="1024x1024",
+ n=1,
+)
+```
+
+## Run through one API gateway
+
+APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
+
+[Create an APIMart account and run this workflow](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-register-47639944) | [API documentation](https://apimart-click-tracker.luyx031226.chatgpt.site/r/gh-apimart-ai-generation-benchmarks-docs-en-quickstart-e3cffa72)
+
+## Review checklist
+
+- Did the output follow the requested composition and protected attributes?
+- Were the model identifier, parameters, and test date recorded?
+- How many retries were required for one accepted output?
+- Does the output contain invented text, marks, or unwanted objects?
+- Can another contributor reproduce the request from the stored fixture?
+
+## Disclosure
+
+This open-source benchmark is maintained by APIMart Labs. Links to APIMart are promotional and directly related to running the documented API examples.
diff --git a/site/comparisons/ai-architecture-rendering-model-comparison/index.html b/site/comparisons/ai-architecture-rendering-model-comparison/index.html
index afcd767..e433828 100644
--- a/site/comparisons/ai-architecture-rendering-model-comparison/index.html
+++ b/site/comparisons/ai-architecture-rendering-model-comparison/index.html
@@ -1 +1 @@
-AI Architecture Rendering Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Architecture Rendering Model Comparison
A blue-hour coastal house prompt tests architectural realism, materials, atmosphere, latency, and cost across three image generation models.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
159.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
43.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · coastal-house-dusk · 16:9
Reproduce the request.
Low modern coastal house at blue hour, weathered cedar and board-formed concrete, dune grasses moving in sea wind, warm interior light, realistic architectural photography, human-scale lens.
OBSERVED FINDING
All three produced convincing architecture. Gemini completed in 32 seconds, while GPT delivered richer detail but required 159 seconds for this fixture.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Architecture Rendering Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Architecture Rendering Model Comparison
A blue-hour coastal house prompt tests architectural realism, materials, atmosphere, latency, and cost across three image generation models.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
159.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
43.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · coastal-house-dusk · 16:9
Reproduce the request.
Low modern coastal house at blue hour, weathered cedar and board-formed concrete, dune grasses moving in sea wind, warm interior light, realistic architectural photography, human-scale lens.
OBSERVED FINDING
All three produced convincing architecture. Gemini completed in 32 seconds, while GPT delivered richer detail but required 159 seconds for this fixture.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-billboard-image-generation-comparison/index.html b/site/comparisons/ai-billboard-image-generation-comparison/index.html
index 8a18154..5879234 100644
--- a/site/comparisons/ai-billboard-image-generation-comparison/index.html
+++ b/site/comparisons/ai-billboard-image-generation-comparison/index.html
@@ -1 +1 @@
-AI Billboard Image Generation Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Billboard Image Generation Comparison
A minimal citrus beverage billboard fixture tested across three image APIs with raw outputs, prompt adherence notes, reported cost, and latency.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
56.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · citrus-billboard · 3:2
Reproduce the request.
Cold citrus sparkling water can with realistic condensation, bright cobalt sky, sharp noon shadow, sliced yuzu suspended around the can, bold minimal billboard composition, no invented lettering.
OBSERVED FINDING
GPT most closely matched the minimal no-lettering direction. Flux invented label marks, while Gemini drifted toward a literal outdoor billboard scene.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Billboard Image Generation Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Billboard Image Generation Comparison
A minimal citrus beverage billboard fixture tested across three image APIs with raw outputs, prompt adherence notes, reported cost, and latency.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
56.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · citrus-billboard · 3:2
Reproduce the request.
Cold citrus sparkling water can with realistic condensation, bright cobalt sky, sharp noon shadow, sliced yuzu suspended around the can, bold minimal billboard composition, no invented lettering.
OBSERVED FINDING
GPT most closely matched the minimal no-lettering direction. Flux invented label marks, while Gemini drifted toward a literal outdoor billboard scene.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-dashboard-ui-generation-comparison/index.html b/site/comparisons/ai-dashboard-ui-generation-comparison/index.html
index 3e8286d..3d18493 100644
--- a/site/comparisons/ai-dashboard-ui-generation-comparison/index.html
+++ b/site/comparisons/ai-dashboard-ui-generation-comparison/index.html
@@ -1 +1 @@
-AI Dashboard UI Generation Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Dashboard UI Generation Comparison
A dense climate intelligence dashboard generated by GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with measured API performance.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
102.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
2
Requested size
1024x1024
SHARED PROMPT · climate-dashboard · 16:10
Reproduce the request.
Editorial analytics dashboard for a climate intelligence product, dense but legible information hierarchy, off-white canvas, dark graphite type, acid green data accents, modular card grid, desktop viewport.
OBSERVED FINDING
GPT produced the most legible information hierarchy. Gemini offered the strongest cost-latency compromise, while Flux contained more pseudo-text.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Dashboard UI Generation Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Dashboard UI Generation Comparison
A dense climate intelligence dashboard generated by GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with measured API performance.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
102.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
2
Requested size
1024x1024
SHARED PROMPT · climate-dashboard · 16:10
Reproduce the request.
Editorial analytics dashboard for a climate intelligence product, dense but legible information hierarchy, off-white canvas, dark graphite type, acid green data accents, modular card grid, desktop viewport.
OBSERVED FINDING
GPT produced the most legible information hierarchy. Gemini offered the strongest cost-latency compromise, while Flux contained more pseudo-text.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-exploded-product-diagram-comparison/index.html b/site/comparisons/ai-exploded-product-diagram-comparison/index.html
index 1a239c5..ae66fa7 100644
--- a/site/comparisons/ai-exploded-product-diagram-comparison/index.html
+++ b/site/comparisons/ai-exploded-product-diagram-comparison/index.html
@@ -1 +1 @@
-AI Exploded Product Diagram Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Exploded Product Diagram Comparison
An exploded running-shoe technical view tests component separation and alignment with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
59.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
43.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · editorial-running-shoe · 3:2
Reproduce the request.
Exploded technical view of a running shoe with upper, laces, foam midsole, plate and outsole separated vertically, clean off-white background, precise alignment, soft studio shadows, no labels.
OBSERVED FINDING
All three followed the exploded layout. GPT and Flux achieved sharper component separation, while Gemini showed less technical detail.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Exploded Product Diagram Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Exploded Product Diagram Comparison
An exploded running-shoe technical view tests component separation and alignment with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
59.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
43.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · editorial-running-shoe · 3:2
Reproduce the request.
Exploded technical view of a running shoe with upper, laces, foam midsole, plate and outsole separated vertically, clean off-white background, precise alignment, soft studio shadows, no labels.
OBSERVED FINDING
All three followed the exploded layout. GPT and Flux achieved sharper component separation, while Gemini showed less technical detail.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-food-photography-model-comparison/index.html b/site/comparisons/ai-food-photography-model-comparison/index.html
index d6fe20b..520ce4d 100644
--- a/site/comparisons/ai-food-photography-model-comparison/index.html
+++ b/site/comparisons/ai-food-photography-model-comparison/index.html
@@ -1 +1 @@
-AI Food Photography Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Food Photography Model Comparison
A sourdough editorial photography prompt compared across GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with raw evidence.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
74.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · artisan-bread-editorial · 4:5
Reproduce the request.
Hand-torn sourdough loaf on a linen-covered oak table, open crumb in sharp detail, soft winter window light, a few flour traces, quiet editorial food photography, warm restrained palette.
OBSERVED FINDING
GPT delivered the strongest crumb detail and editorial styling, Flux created a convincing simpler composition, and Gemini placed less emphasis on the food subject.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Food Photography Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Food Photography Model Comparison
A sourdough editorial photography prompt compared across GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with raw evidence.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
74.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · artisan-bread-editorial · 4:5
Reproduce the request.
Hand-torn sourdough loaf on a linen-covered oak table, open crumb in sharp detail, soft winter window light, a few flour traces, quiet editorial food photography, warm restrained palette.
OBSERVED FINDING
GPT delivered the strongest crumb detail and editorial styling, Flux created a convincing simpler composition, and Gemini placed less emphasis on the food subject.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-poster-typography-model-comparison/index.html b/site/comparisons/ai-poster-typography-model-comparison/index.html
index 10076a9..783d1f1 100644
--- a/site/comparisons/ai-poster-typography-model-comparison/index.html
+++ b/site/comparisons/ai-poster-typography-model-comparison/index.html
@@ -1 +1 @@
-AI Poster Typography Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Poster Typography Model Comparison
An experimental music poster prompt tests layout, blank type placeholders, halftone texture, and instruction following across three image generation models.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
68.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
2
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · music-festival-poster · 2:3
Reproduce the request.
Experimental electronic music festival poster, oversized condensed type placeholders, electric blue field, silver halftone sphere, asymmetric Swiss grid, high contrast screen-print texture, leave all wording blank.
OBSERVED FINDING
GPT and Flux followed the blank-placeholder instruction more closely. Gemini rendered invented festival wording despite the prompt explicitly asking for blank type.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Poster Typography Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Poster Typography Model Comparison
An experimental music poster prompt tests layout, blank type placeholders, halftone texture, and instruction following across three image generation models.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
68.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
2
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · music-festival-poster · 2:3
Reproduce the request.
Experimental electronic music festival poster, oversized condensed type placeholders, electric blue field, silver halftone sphere, asymmetric Swiss grid, high contrast screen-print texture, leave all wording blank.
OBSERVED FINDING
GPT and Flux followed the blank-placeholder instruction more closely. Gemini rendered invented festival wording despite the prompt explicitly asking for blank type.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/ai-running-shoe-ad-model-comparison/index.html b/site/comparisons/ai-running-shoe-ad-model-comparison/index.html
index 706306f..efdf9e2 100644
--- a/site/comparisons/ai-running-shoe-ad-model-comparison/index.html
+++ b/site/comparisons/ai-running-shoe-ad-model-comparison/index.html
@@ -1 +1 @@
-AI Running Shoe Ad Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Running Shoe Ad Model Comparison
The same commercial running-shoe motion prompt tested with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview, including latency and cost.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
73.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
22.0 s
Reported cost
$0.0700
Attempts
2
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · running-shoe-motion · 4:5
Reproduce the request.
Technical running shoe frozen above a red clay track at impact, fine dust particles suspended in air, directional hard sunlight, visible outsole detail, energetic diagonal composition, commercial sports photography.
OBSERVED FINDING
GPT created the most polished commercial crop, Flux showed the strongest outsole and track realism, while Gemini removed the athlete and emphasized a floating product shot.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+AI Running Shoe Ad Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
AI Running Shoe Ad Model Comparison
The same commercial running-shoe motion prompt tested with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview, including latency and cost.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
73.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
22.0 s
Reported cost
$0.0700
Attempts
2
Requested size
1024x1024
Gemini 2.5 Flash
Latency
33.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · running-shoe-motion · 4:5
Reproduce the request.
Technical running shoe frozen above a red clay track at impact, fine dust particles suspended in air, directional hard sunlight, visible outsole detail, energetic diagonal composition, commercial sports photography.
OBSERVED FINDING
GPT created the most polished commercial crop, Flux showed the strongest outsole and track realism, while Gemini removed the athlete and emphasized a floating product shot.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/consistent-character-ai-model-comparison/index.html b/site/comparisons/consistent-character-ai-model-comparison/index.html
index 63c79bb..b10b781 100644
--- a/site/comparisons/consistent-character-ai-model-comparison/index.html
+++ b/site/comparisons/consistent-character-ai-model-comparison/index.html
@@ -1 +1 @@
-Consistent Character AI Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
Consistent Character AI Model Comparison
A three-view forest ranger character sheet generated with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview using one shared prompt.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
79.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · forest-ranger-sheet · 3:2
Reproduce the request.
Character sheet for Mara, a forest ranger with cropped auburn hair, olive field jacket, brass compass and worn canvas backpack; front, profile and three-quarter views; neutral studio background; consistent facial features.
OBSERVED FINDING
All three maintained the ranger identity across views. GPT added the richest close-up and prop detail; Flux and Gemini produced cleaner, simpler turnaround sheets.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+Consistent Character AI Model Comparison - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
Consistent Character AI Model Comparison
A three-view forest ranger character sheet generated with GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview using one shared prompt.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
79.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · forest-ranger-sheet · 3:2
Reproduce the request.
Character sheet for Mara, a forest ranger with cropped auburn hair, olive field jacket, brass compass and worn canvas backpack; front, profile and three-quarter views; neutral studio background; consistent facial features.
OBSERVED FINDING
All three maintained the ranger identity across views. GPT added the richest close-up and prop detail; Flux and Gemini produced cleaner, simpler turnaround sheets.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/gpt-image-vs-flux-vs-gemini-product-photography/index.html b/site/comparisons/gpt-image-vs-flux-vs-gemini-product-photography/index.html
index 534190f..ec01e06 100644
--- a/site/comparisons/gpt-image-vs-flux-vs-gemini-product-photography/index.html
+++ b/site/comparisons/gpt-image-vs-flux-vs-gemini-product-photography/index.html
@@ -1 +1 @@
-GPT Image vs Flux vs Gemini: Product Photography Test - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
GPT Image vs Flux vs Gemini: Product Photography Test
A reproducible product photography comparison using the same ceramic skincare prompt across GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
49.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
34.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · ceramic-skincare-hero · 1:1
Reproduce the request.
Editorial product photograph of a matte ceramic skincare bottle on pale limestone, soft north-window light, delicate contact shadow, warm neutral palette, 85mm lens compression, no text, no logo.
OBSERVED FINDING
GPT Image 2 delivered the strongest editorial staging, Flux followed the literal window-light direction most closely, and Gemini produced the cleanest but least distinctive pack shot.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+GPT Image vs Flux vs Gemini: Product Photography Test - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
GPT Image vs Flux vs Gemini: Product Photography Test
A reproducible product photography comparison using the same ceramic skincare prompt across GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
49.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
34.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · ceramic-skincare-hero · 1:1
Reproduce the request.
Editorial product photograph of a matte ceramic skincare bottle on pale limestone, soft north-window light, delicate contact shadow, warm neutral palette, 85mm lens compression, no text, no logo.
OBSERVED FINDING
GPT Image 2 delivered the strongest editorial staging, Flux followed the literal window-light direction most closely, and Gemini produced the cleanest but least distinctive pack shot.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
diff --git a/site/comparisons/transparent-background-ai-image-api-test/index.html b/site/comparisons/transparent-background-ai-image-api-test/index.html
index 228627e..b3c8434 100644
--- a/site/comparisons/transparent-background-ai-image-api-test/index.html
+++ b/site/comparisons/transparent-background-ai-image-api-test/index.html
@@ -1 +1 @@
-Transparent Background AI Image API Test - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
Transparent Background AI Image API Test
A transparent product cutout test comparing GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with pixel-format verification.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
74.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · transparent-headphones · 1:1
Reproduce the request.
Over-ear headphones isolated as a clean ecommerce cutout, transparent background, accurate soft edge detail around fabric ear pads, neutral product color, no floor, no text, no added accessories.
OBSERVED FINDING
All three outputs were opaque 24-bit RGB PNGs. GPT and Flux drew checkerboards and Gemini used a dark background, so none satisfied the true-transparency requirement.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
\ No newline at end of file
+Transparent Background AI Image API Test - APIMart LabsMEASURED 2026-07-21 · 3 MODEL API TEST
Transparent Background AI Image API Test
A transparent product cutout test comparing GPT Image 2, Flux 2 Flex, and Gemini 2.5 Flash Image Preview with pixel-format verification.
One API gateway for multiple AI models.
APIMart is a multi-model AI API gateway with an OpenAI-compatible endpoint. Use one account to access supported image models, compare current pricing, and choose a cost-efficient option for each workload.
One accepted output per model. Cost and latency are copied from the stored run records; this dated snapshot is not a universal ranking.
GPT Image 2
Latency
74.0 s
Reported cost
$0.0085
Attempts
1
Requested size
1024x1024
Flux 2 Flex
Latency
32.0 s
Reported cost
$0.0700
Attempts
1
Requested size
1024x1024
Gemini 2.5 Flash
Latency
32.0 s
Reported cost
$0.0125
Attempts
1
Requested size
1024x1024
SHARED PROMPT · transparent-headphones · 1:1
Reproduce the request.
Over-ear headphones isolated as a clean ecommerce cutout, transparent background, accurate soft edge detail around fabric ear pads, neutral product color, no floor, no text, no added accessories.
OBSERVED FINDING
All three outputs were opaque 24-bit RGB PNGs. GPT and Flux drew checkerboards and Gemini used a dark background, so none satisfied the true-transparency requirement.
Test date: 2026-07-21. The report uses one accepted output from each API and preserves the original images and JSON run records in GitHub. Availability, behavior, and pricing can change.
