feat(together-ai): add GLM-5.1, Kimi K2.6, DeepSeek V4 Pro - #2094
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
WalkthroughAdds multiple Together AI provider entries across model catalogs (deepseek, moonshot, zai, openai) and marks several Meta provider variants as deactivated by adding Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Also deactivate gemma-2-27b-it on Together AI as of 2026-04-25. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Mark llama-3.1-8b-instruct and llama-4-scout providers as deactivated 2026-04-25 across providers without an existing date. Adjust together-ai entries: omit reasoning output for DeepSeek V4 Pro (not exposed) and skip e2e for Kimi K2.6 (no streaming content). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
af7405d to
ec080cb
Compare
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: af7405de4e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| requestPrice: 0, | ||
| contextSize: 262144, | ||
| maxOutput: 32768, | ||
| streaming: true, |
There was a problem hiding this comment.
This mapping advertises streaming: true, so apps/gateway/src/chat/chat.ts will allow streamed requests for it, but this same provider entry is already tagged test: "skip" due to known streaming failures (no streaming content returned in the commit context). As shipped, clients using stream: true can be routed to a configuration that returns empty streamed responses instead of being rejected early; setting streaming: false (or deactivating the mapping until fixed) would prevent broken runtime behavior.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Pull request overview
Adds/updates Together AI provider mappings for several newly supported models and deactivates a set of provider mappings that have been retired.
Changes:
- Add Together AI mappings (with pricing/capabilities) for GLM-5.1, Kimi K2.6, and DeepSeek V4 Pro.
- Mark several existing provider mappings as deactivated as of 2026-04-25 (Gemma 2 27B IT on Together AI; multiple Llama 3.1 8B Instruct mappings; Together AI’s Llama 4 Scout mapping).
- Apply provider-specific behavior tweaks (e.g., omit reasoning output for Together AI DeepSeek V4 Pro; skip e2e for Together AI Kimi K2.6).
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| packages/models/src/models/zai.ts | Adds Together AI mapping for GLM-5.1 with pricing and capability flags. |
| packages/models/src/models/moonshot.ts | Adds Together AI mapping for Kimi K2.6 (tests skipped). |
| packages/models/src/models/deepseek.ts | Adds Together AI mapping for DeepSeek V4 Pro, including cached pricing and reasoning-output behavior. |
| packages/models/src/models/google.ts | Marks Together AI Gemma 2 27B IT mapping as deactivated. |
| packages/models/src/models/meta.ts | Deactivates remaining Llama 3.1 8B Instruct mappings and Together AI’s Llama 4 Scout mapping. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| streaming: true, | ||
| vision: false, | ||
| tools: true, | ||
| jsonOutput: false, | ||
| deactivatedAt: new Date("2026-04-25"), |
There was a problem hiding this comment.
For the llama-4-scout model’s Together AI mapping, the configured modelName in this provider entry appears to point to a Llama 3.1 405B Turbo model rather than a Llama 4 Scout variant. Even though this mapping is now deactivated, this mismatch is confusing and can lead to incorrect routing/analytics if the mapping is ever referenced; please verify the intended Together model slug and either correct modelName (and any related pricing/context fields) or rename/remove the model definition so the id/name and modelName align.
JSON output mode is unreliable on Together AI for both, so disabled. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 52eed64412
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| vision: false, | ||
| tools: true, | ||
| jsonOutput: false, | ||
| deactivatedAt: new Date("2026-04-25"), |
There was a problem hiding this comment.
This deactivatedAt change is applied to the llama-4-scout provider record whose modelName is meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo, so after 2026-04-25 requests targeting together-ai/meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo will be rejected as deactivated. That disables the 3.1-405B mapping rather than the actual Scout model mapping (meta-llama/llama-4-scout-17b-16e-instruct), so the retirement is being enforced on the wrong entry.
Useful? React with 👍 / 👎.
Summary
gemma-2-27b-iton Together AI as deactivated 2026-04-25.llama-3.1-8b-instructprovider mappings (aws-bedrock, nebius, inference.net, cerebras, novita) and the solellama-4-scoutmapping (together-ai) as of 2026-04-25 — together-ai forllama-3.1-8b-instructalready had 2026-03-27.reasoningOutput: "omit"for DeepSeek V4 Pro (Together AI doesn't expose reasoning),test: "skip"for Kimi K2.6 (no streaming content returned).Together AI pricing
Test plan
pnpm --filter @llmgateway/models buildTEST_MODELS="together-ai/glm-5.1,together-ai/deepseek-v4-pro" pnpm test:e2e— all passtest: "skip"🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Updates