@juspay/neurolink
Version:
The pipe layer of an AI nervous system: one interface connecting provider neurons to your application, across three inference types — generate, stream and decide. `decide` returns typed, calibrated judgements from a non-generative model (~400ms, ~$0.00002
177 lines (176 loc) • 6.99 kB
JSON
{
"$schema": "./provider-catalog.schema.json",
"id": "groq",
"displayName": "Groq",
"aliases": [],
"tier": 2,
"wire": {
"baseURL": "https://api.groq.com/openai/v1"
},
"models": {
"default": "openai/gpt-oss-120b",
"fallbacks": [
"openai/gpt-oss-120b",
"openai/gpt-oss-20b",
"qwen/qwen3.8-27b",
"groq/compound"
],
"defaultContextWindow": 128000,
"defaultMaxOutputTokens": 4096,
"catalog": {
"openai/gpt-oss-120b": {
"enumMember": "GPT_OSS_120B",
"contextWindow": 131072,
"vision": false,
"status": "production",
"description": "Recommended - GPT-OSS 120B, Groq's flagship open model"
},
"openai/gpt-oss-20b": {
"enumMember": "GPT_OSS_20B",
"contextWindow": 131072,
"vision": false,
"status": "production",
"description": "GPT-OSS 20B — lower latency, smaller model"
},
"qwen/qwen3.8-27b": {
"enumMember": "QWEN_3_8_27B",
"contextWindow": 131072,
"vision": true,
"status": "production",
"description": "Qwen 3.8 27B — Groq's only vision-capable model"
},
"qwen/qwen3.6-27b": {
"enumMember": "QWEN_3_6_27B",
"contextWindow": 131072,
"vision": false,
"status": "production",
"description": "Qwen 3.6 27B"
},
"groq/compound": {
"enumMember": "COMPOUND",
"contextWindow": 131072,
"vision": false,
"status": "production",
"description": "Groq Compound — agentic system with built-in tool use"
},
"groq/compound-mini": {
"enumMember": "COMPOUND_MINI",
"contextWindow": 131072,
"vision": false,
"status": "production",
"description": "Groq Compound Mini — lighter agentic system"
},
"allam-2-7b": {
"enumMember": "ALLAM_2_7B",
"contextWindow": 4096,
"vision": false,
"status": "production",
"description": "ALLaM 2 7B — Arabic-focused model"
},
"llama-3.3-70b-versatile": {
"contextWindow": 131072,
"pricingPerMTok": { "input": 0.59, "output": 0.79 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq returns 404 'does not exist'"
},
"llama-3.1-8b-instant": {
"contextWindow": 128000,
"pricingPerMTok": { "input": 0.05, "output": 0.08 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq returns 404 'does not exist'"
},
"gemma2-9b-it": {
"enumMember": "GEMMA_2_9B_IT",
"contextWindow": 8192,
"pricingPerMTok": { "input": 0.2, "output": 0.2 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq reports 'decommissioned'"
},
"mixtral-8x7b-32768": {
"contextWindow": 32768,
"pricingPerMTok": { "input": 0.24, "output": 0.24 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq reports 'decommissioned'"
},
"llama-guard-3-8b": {
"contextWindow": 8192,
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq reports 'decommissioned'"
},
"llama-3.2-90b-vision-preview": {
"contextWindow": 128000,
"pricingPerMTok": { "input": 0.9, "output": 0.9 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq reports 'decommissioned'"
},
"llama-3.2-11b-vision-preview": {
"contextWindow": 128000,
"pricingPerMTok": { "input": 0.18, "output": 0.18 },
"vision": false,
"status": "retired",
"description": "Retired 2026-08 — Groq reports 'decommissioned'"
}
},
"visionModel": "qwen/qwen3.8-27b",
"topModels": [
"openai/gpt-oss-120b",
"openai/gpt-oss-20b",
"qwen/qwen3.8-27b",
"groq/compound"
]
},
"capabilities": {
"text": true,
"streaming": true,
"tools": true,
"toolsWithStreaming": true,
"structuredOutput": true,
"structuredOutputWithTools": false,
"embeddings": false,
"thinking": false
},
"errorRules": [
{
"status": 401,
"pattern": "Invalid API key|Authentication|invalid_api_key",
"class": "authentication",
"message": "Invalid Groq API key. Check {apiKeyEnvVar}. Get one at https://console.groq.com/keys"
},
{
"pattern": "model_decommissioned",
"class": "invalid-model",
"message": "Groq model '{model}' was decommissioned. Pick a current model from https://console.groq.com/docs/models."
}
],
"quirks": {
"timeoutErrorClass": "provider"
},
"setup": {
"url": "https://console.groq.com/keys",
"apiKeyFormat": null,
"billingPolicy": "free-tier",
"instructions": [
"1. Visit: https://console.groq.com/keys",
"2. Sign in to your Groq account",
"3. Create a new API key",
"4. Set {apiKeyEnvVar} in your .env file"
]
},
"evidence": {
"rosterVerified": {
"date": "2026-08-30",
"method": "GET /models against the live account, cross-checked with a real chat call per id"
},
"liveMatrix": {
"date": "2026-08-30",
"result": "All 7 previously-catalogued ids are gone: Groq answers 404 'does not exist' for the llama ids and 'has been decommissioned' for gemma2/mixtral/llama-guard/llama-3.2-vision. The 7 replacements listed here each returned 200 to a real chat call. Vision was re-tested with a valid 64x64 PNG: qwen/qwen3.8-27b describes the image correctly and is the only vision-capable id; gpt-oss-120b and groq/compound reject an image part with 'messages[0].content must be a string'."
},
"addedInPR": "https://github.com/juspay/neurolink/pull/1587"
}
}