27 models across OpenAI, Anthropic, Google/Vertex, Azure, AWS Bedrock, xAI, Groq, DeepInfra and Mistral — one endpoint, individually quality-scored per task so routing has real price spread to choose from. Prices are NeuroRoute's own metered platform rate per million tokens; BYOK customers pay providers directly at their published rates.
| Model | Provider | Quality | $/1M in | $/1M out | Context |
|---|---|---|---|---|---|
| Claude Haiku 4.5 | Anthropic | 88 | $1 | $5 | 200K |
| Claude Opus 4.8 | Anthropic | 98 | $5 | $25 | 200K |
| Claude Sonnet 4.6 | Anthropic | 95 | $3 | $15 | 200K |
| GPT-4.1 (Azure) | Azure OpenAI | 96 | $2 | $8 | 1M |
| GPT-4o (Azure) | Azure OpenAI | 93 | $2.5 | $10 | 128K |
| Claude Opus 4.8 (Bedrock) | Bedrock | 93 | $5 | $25 | 200K |
| Claude Sonnet 4.6 (Bedrock) | Bedrock | 90 | $3 | $15 | 200K |
| Llama 3.3 70B (Bedrock) | Bedrock | 84 | $0.72 | $0.72 | 128K |
| Nova Pro (Bedrock) | Bedrock | 83 | $0.8 | $3.2 | 300K |
| DeepSeek V3 | DeepInfra | 88 | $0.32 | $0.89 | 128K |
| GLM 5.2 | DeepInfra | 92 | $0.95 | $3 | 1M |
| Kimi K2.7 Code | DeepInfra | 85 | $0.74 | $3.5 | 256K |
| Llama 3.3 70B Turbo | DeepInfra | 84 | $0.1 | $0.32 | 128K |
| Qwen3 32B | DeepInfra | 83 | $0.08 | $0.28 | 128K |
| Gemini 2.5 Flash | 86 | $0.3 | $2.5 | 1M | |
| Gemini 2.5 Pro | 94 | $1.25 | $10 | 2M | |
| Llama 3.1 8B | Groq | 68 | $0.05 | $0.08 | 128K |
| Llama 3.3 70B | Groq | 82 | $0.59 | $0.79 | 128K |
| Mistral Large | Mistral | 88 | $2 | $6 | 128K |
| GPT-4.1 | OpenAI | 96 | $2 | $8 | 1M |
| GPT-4o | OpenAI | 93 | $2.5 | $10 | 128K |
| GPT-4o-mini | OpenAI | 81 | $0.15 | $0.6 | 128K |
| Claude Opus 5 (Vertex) | Vertex | 97 | $5 | $25 | 1M |
| Gemini 3.1 Flash-Lite (Vertex) | Vertex | 80 | $0.04 | $0.15 | 1M |
| Gemini 3.5 Flash (Vertex) | Vertex | 90 | $0.075 | $0.3 | 1M |
| Gemma 3 27B (Vertex) | Vertex | 79 | $0.1 | $0.3 | 128K |
| Grok 3 | xAI | 92 | $3 | $15 | 131K |
Quality is NeuroRoute's own per-task score (0–100), retrained hourly from real feedback — not a third-party benchmark. For the live, authoritative catalog at any moment, query GET /v1/models directly.