Skip to content

feat: add GLM-4.6-turbo model to Chutes provider #8514

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 2 commits into from
Closed

feat: add GLM-4.6-turbo model to Chutes provider #8514

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6c83f94
feat: add GLM-4.6-turbo model to Chutes provider
raulcorreia7 Oct 5, 2025
98c95d2
fix: update GLM-4.6-FP8 pricing to match API values
raulcorreia7 Oct 5, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 23 additions & 8 deletions packages/types/src/providers/chutes.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@ export type ChutesModelId =
| "zai-org/GLM-4.5-FP8"
| "zai-org/GLM-4.5-turbo"
| "zai-org/GLM-4.6-FP8"
| "zai-org/GLM-4.6-turbo"
| "moonshotai/Kimi-K2-Instruct-75k"
| "moonshotai/Kimi-K2-Instruct-0905"
| "Qwen/Qwen3-235B-A22B-Thinking-2507"
Expand Down Expand Up @@ -87,7 +88,8 @@ export const chutesModels = {
supportsPromptCache: false,
inputPrice: 0.23,
outputPrice: 0.9,
description: "DeepSeek‑V3.1‑Terminus is an update to V3.1 that improves language consistency by reducing CN/EN mix‑ups and eliminating random characters, while strengthening agent capabilities with notably better Code Agent and Search Agent performance.",
description:
"DeepSeek‑V3.1‑Terminus is an update to V3.1 that improves language consistency by reducing CN/EN mix‑ups and eliminating random characters, while strengthening agent capabilities with notably better Code Agent and Search Agent performance.",
},
"deepseek-ai/DeepSeek-V3.1-turbo": {
maxTokens: 32768,
Expand All @@ -96,7 +98,8 @@ export const chutesModels = {
supportsPromptCache: false,
inputPrice: 1.0,
outputPrice: 3.0,
description: "DeepSeek-V3.1-turbo is an FP8, speculative-decoding turbo variant optimized for ultra-fast single-shot queries (~200 TPS), with outputs close to the originals and solid function calling/reasoning/structured output, priced at $1/M input and $3/M output tokens, using 2× quota per request and not intended for bulk workloads.",
description:
"DeepSeek-V3.1-turbo is an FP8, speculative-decoding turbo variant optimized for ultra-fast single-shot queries (~200 TPS), with outputs close to the originals and solid function calling/reasoning/structured output, priced at $1/M input and $3/M output tokens, using 2× quota per request and not intended for bulk workloads.",
},
"deepseek-ai/DeepSeek-V3.2-Exp": {
maxTokens: 163840,
Expand All @@ -105,7 +108,8 @@ export const chutesModels = {
supportsPromptCache: false,
inputPrice: 0.25,
outputPrice: 0.35,
description: "DeepSeek-V3.2-Exp is an experimental LLM that introduces DeepSeek Sparse Attention to improve long‑context training and inference efficiency while maintaining performance comparable to V3.1‑Terminus.",
description:
"DeepSeek-V3.2-Exp is an experimental LLM that introduces DeepSeek Sparse Attention to improve long‑context training and inference efficiency while maintaining performance comparable to V3.1‑Terminus.",
},
"unsloth/Llama-3.3-70B-Instruct": {
maxTokens: 32768, // From Groq
Expand Down Expand Up @@ -321,11 +325,21 @@ export const chutesModels = {
contextWindow: 202752,
supportsImages: false,
supportsPromptCache: false,
inputPrice: 0,
outputPrice: 0,
inputPrice: 0.5,
outputPrice: 1.75,
description:
"GLM-4.6 introduces major upgrades over GLM-4.5, including a longer 200K-token context window for complex tasks, stronger coding performance in benchmarks and real-world tools (such as Claude Code, Cline, Roo Code, and Kilo Code), improved reasoning with tool use during inference, more capable and efficient agent integration, and refined writing that better matches human style, readability, and natural role-play scenarios.",
},
"zai-org/GLM-4.6-turbo": {
Copy link
Contributor

@roomote roomote bot Oct 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: Add/update tests to cover this new model ID in the Chutes provider. Include assertions that the model appears in the supported IDs and that its ModelInfo (contextWindow, maxTokens, pricing, image/prompt-cache support) matches expectations, e.g., in src/api/providers/tests/chutes.spec.ts.

maxTokens: 32768,
contextWindow: 202752,
supportsImages: false,
supportsPromptCache: false,
inputPrice: 1.15,
outputPrice: 3.25,
description:
"GLM-4.6-turbo model with 200K+ token context window, optimized for fast inference with enhanced reasoning and tool use capabilities.",
Comment on lines +340 to +341
Copy link
Contributor

@roomote roomote bot Oct 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P3: Description says "200K+" while contextWindow is 202752. For consistency and clarity, consider using the exact number in the description (e.g., "202,752-token context window").

},
"Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": {
maxTokens: 32768,
contextWindow: 262144,
Expand Down Expand Up @@ -387,8 +401,9 @@ export const chutesModels = {
contextWindow: 262144,
supportsImages: true,
supportsPromptCache: false,
inputPrice: 0.1600,
outputPrice: 0.6500,
description: "Qwen3‑VL‑235B‑A22B‑Thinking is an open‑weight MoE vision‑language model (235B total, ~22B activated) optimized for deliberate multi‑step reasoning with strong text‑image‑video understanding and long‑context capabilities.",
inputPrice: 0.16,
outputPrice: 0.65,
description:
"Qwen3‑VL‑235B‑A22B‑Thinking is an open‑weight MoE vision‑language model (235B total, ~22B activated) optimized for deliberate multi‑step reasoning with strong text‑image‑video understanding and long‑context capabilities.",
},
} as const satisfies Record<string, ModelInfo>