chore(llma): Update LLM costs

[scheduled-actions-posthog]

This is an automated PR to update LLM pricing models.

Reviewer checklist

Please verify that there aren't any odd changes in the pricing by checking against the official pricing pages:

• Anthropic Pricing
• OpenAI Pricing
• Google AI (Gemini) Pricing
• Mistral AI Pricing

[greptile-apps]

Comments Outside Diff (1)

1. nodejs/src/ingestion/ai/costs/providers/llm-costs.json, line 5895-5906 (link)

   Suspicious default pricing for qwen3-235b-a22b-thinking-2507

   The default completion_token cost dropped ~15× from 0.000001495 ($1.495/M) to 1e-7 ($0.10/M), and prompt, completion, and cache_read tokens are now all priced identically at 1e-7. OpenRouter's own pricing page currently shows $0.1495/M input and $1.495/M output for this model — the previous values. Having the same unit cost for prompt and completion tokens is unusual for any model, and particularly surprising for a dedicated thinking/reasoning variant that typically generates far more output tokens than input. If these values were pulled from an upstream API change, it would be worth cross-checking the OpenRouter page directly before merging, since an underpriced completion token could significantly skew cost attribution.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: nodejs/src/ingestion/ai/costs/providers/llm-costs.json
   Line: 5895-5906
   
   Comment:
   **Suspicious default pricing for `qwen3-235b-a22b-thinking-2507`**
   
   The `default` completion_token cost dropped ~15× from `0.000001495` ($1.495/M) to `1e-7` ($0.10/M), and prompt, completion, and cache_read tokens are now all priced identically at `1e-7`. OpenRouter's own pricing page currently shows $0.1495/M input and **$1.495/M output** for this model — the previous values. Having the same unit cost for prompt and completion tokens is unusual for any model, and particularly surprising for a dedicated thinking/reasoning variant that typically generates far more output tokens than input. If these values were pulled from an upstream API change, it would be worth cross-checking the OpenRouter page directly before merging, since an underpriced completion token could significantly skew cost attribution.
   
   How can I resolve this? If you propose a fix, please make it concise.

   Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Prompt To Fix All With AI

Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 1
nodejs/src/ingestion/ai/costs/providers/llm-costs.json:5895-5906
**Suspicious default pricing for `qwen3-235b-a22b-thinking-2507`**

The `default` completion_token cost dropped ~15× from `0.000001495` ($1.495/M) to `1e-7` ($0.10/M), and prompt, completion, and cache_read tokens are now all priced identically at `1e-7`. OpenRouter's own pricing page currently shows $0.1495/M input and **$1.495/M output** for this model — the previous values. Having the same unit cost for prompt and completion tokens is unusual for any model, and particularly surprising for a dedicated thinking/reasoning variant that typically generates far more output tokens than input. If these values were pulled from an upstream API change, it would be worth cross-checking the OpenRouter page directly before merging, since an underpriced completion token could significantly skew cost attribution.

_{Reviews (1): Last reviewed commit: "chore(llma): update LLM costs" | Re-trigger Greptile}

[Radu-Raicea]

@PostHog/team-ai-gateway

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-06-01 13:39 UTC	Run
prod-us	✅ Deployed	2026-06-01 13:57 UTC	Run
prod-eu	✅ Deployed	2026-06-01 14:01 UTC	Run