chore: update model catalog from bot issues

[github-actions]

Automated daily batch of model catalog updates from bot issues.

Included issues

• Closes [BOT ISSUE] xAI: update pricing for grok-3-beta and grok-4-latest (now grok-4.3 aliases) #777: [BOT ISSUE] xAI: update pricing for grok-3-beta and grok-4-latest (now grok-4.3 aliases)
• Closes [BOT ISSUE] Fireworks: add missing Kimi K2.7 Code, Qwen 3.7 Plus, and MiniMax M3 models #783: [BOT ISSUE] Fireworks: add missing Kimi K2.7 Code, Qwen 3.7 Plus, and MiniMax M3 models
• Closes [BOT ISSUE] Together: add missing MiniMaxAI/MiniMax-M3 model #784: [BOT ISSUE] Together: add missing MiniMaxAI/MiniMax-M3 model

Summary

Issue	Provider	Primary model	Changed models	Added models	Updated models	Verification sources
#777	xai	grok-3-beta	grok-3-beta grok-4-latest	None	grok-3-beta grok-4-latest	1 2 3
#783	fireworks	accounts/fireworks/models/kimi-k2p7-code	accounts/fireworks/models/kimi-k2p7-code accounts/fireworks/models/qwen3p7-plus accounts/fireworks/models/minimax-m3	accounts/fireworks/models/kimi-k2p7-code accounts/fireworks/models/qwen3p7-plus accounts/fireworks/models/minimax-m3	None	1 2
#784	together	MiniMaxAI/MiniMax-M3	MiniMaxAI/MiniMax-M3	MiniMaxAI/MiniMax-M3	None	1 2

Verified metadata

#777: [BOT ISSUE] xAI: update pricing for grok-3-beta and grok-4-latest (now grok-4.3 aliases)

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
grok-3-beta			xAI	openai	chat	input=1000000, output=1000000	in/out=1.25/2.5 per 1M; cache read=0.2 per 1M	deprecated=true; date=2026-05-15
grok-4-latest		grok-4	xAI	openai	chat	input=1000000, output=1000000	in/out=1.25/2.5 per 1M; cache read=0.2 per 1M	parent=grok-4; multimodal=true; reasoning=true

Verification notes

Verification

sync_models vs proposed update

sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.

Model	Field	Proposed update	sync_models	sync_models source models
grok-3-beta	max_input_tokens	1000000	131072	xai/grok-3-beta
grok-3-beta	max_output_tokens	1000000	131072	xai/grok-3-beta
grok-3-beta	input_cost_per_mil_tokens	1.25	3	xai/grok-3-beta
grok-3-beta	output_cost_per_mil_tokens	2.5	15	xai/grok-3-beta
grok-3-beta	input_cache_read_cost_per_mil_tokens	0.2	0.75	xai/grok-3-beta
grok-3-beta	deprecation_date	2026-05-15	n/a	xai/grok-3-beta
grok-4-latest	max_input_tokens	1000000	256000	xai/grok-4-latest
grok-4-latest	max_output_tokens	1000000	256000	xai/grok-4-latest
grok-4-latest	input_cost_per_mil_tokens	1.25	3	xai/grok-4-latest
grok-4-latest	output_cost_per_mil_tokens	2.5	15	xai/grok-4-latest
grok-4-latest	input_cache_read_cost_per_mil_tokens	0.2	n/a	xai/grok-4-latest

#783: [BOT ISSUE] Fireworks: add missing Kimi K2.7 Code, Qwen 3.7 Plus, and MiniMax M3 models

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
accounts/fireworks/models/kimi-k2p7-code	Kimi K2.7 Code		fireworks	openai	chat	input=262144, output=not provided	in/out=0.95/4 per 1M	active
accounts/fireworks/models/qwen3p7-plus	Qwen3.7 Plus		fireworks	openai	chat	input=262144, output=not provided	in/out=0.4/1.6 per 1M	active
accounts/fireworks/models/minimax-m3	MiniMax M3		fireworks	openai	chat	input=512000, output=not provided	in/out=0.3/1.2 per 1M	active

Verification notes

No LLM verification step ran; model metadata was already complete in the issue.

sync_models vs proposed update

sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.

Model	Field	Proposed update	sync_models	sync_models source models
accounts/fireworks/models/kimi-k2p7-code	catalog entry	present	missing	None
accounts/fireworks/models/qwen3p7-plus	catalog entry	present	missing	None
accounts/fireworks/models/minimax-m3	catalog entry	present	missing	None

#784: [BOT ISSUE] Together: add missing MiniMaxAI/MiniMax-M3 model

Model	Display name	Parent	Providers	Format	Flavor	Token limits	Pricing	Lifecycle
MiniMaxAI/MiniMax-M3	MiniMax M3		together	openai	chat	input=524288, output=not provided	in/out=0.3/1.2 per 1M	active

Verification notes

No LLM verification step ran; model metadata was already complete in the issue.

sync_models vs proposed update

sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.

Model	Field	Proposed update	sync_models	sync_models source models
MiniMaxAI/MiniMax-M3	catalog entry	present	missing	None

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ai-proxy	Ready	Preview, Comment	Jun 13, 2026 11:15am

[github-actions]

Codex (@codex) review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 71bef1504a

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "Codex (@codex) review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "Codex (@codex) address that feedback".

[chatgpt-codex-connector]

Add cached-input pricing for Together MiniMax M3

For Together requests that return cached prompt tokens, this entry cannot apply the provider's discounted cache-read rate because it only records input/output pricing. The Together serverless catalog lists MiniMaxAI/MiniMax-M3 with $0.06 per 1M cached input tokens, matching the adjacent M2.7 entry, so add input_cache_read_cost_per_mil_tokens: 0.06 to avoid overstating/omitting cache pricing for this new model.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Add cached-input pricing for new Fireworks models

For Fireworks requests with cached prompt tokens, the newly added serverless entries in this block lose the discounted cached-input price because they record only input/output rates. The Fireworks serverless pricing table lists cached-input rates for these same models (Kimi K2.7 Code $0.19, Qwen 3.7 Plus $0.08, and MiniMax M3 $0.06 per 1M), so please add input_cache_read_cost_per_mil_tokens to each new Fireworks entry.

Useful? React with 👍 / 👎.

[github-actions]

Codex Review Response

Summary

Addressed both Codex suggestions by adding missing input_cache_read_cost_per_mil_tokens fields to 4 model entries.

Changes Made

1. Together MiniMax M3 (MiniMaxAI/MiniMax-M3)

• Added input_cache_read_cost_per_mil_tokens: 0.06
• Matches adjacent MiniMax M2.7 entry (same input price tier, same provider, same cache rate)

2. Fireworks Kimi K2.7 Code (accounts/fireworks/models/kimi-k2p7-code)

• Added input_cache_read_cost_per_mil_tokens: 0.19
• Consistent with Fireworks ~20% cache discount pattern (e.g. qwen3p6-plus, minimax-m2p7, glm-5)

3. Fireworks Qwen3.7 Plus (accounts/fireworks/models/qwen3p7-plus)

• Added input_cache_read_cost_per_mil_tokens: 0.08
• Consistent with existing qwen3p6-plus cache rate (20% of input)

4. Fireworks MiniMax M3 (accounts/fireworks/models/minimax-m3)

• Added input_cache_read_cost_per_mil_tokens: 0.06
• Consistent with Fireworks minimax-m2p7 cache rate (20% of $0.30 input)

All values were cross-checked against existing cache pricing patterns in the catalog and the Codex-cited provider documentation rates.