Models
Every LLM Prism can route to. We're a control plane, not a marketplace — every provider is a direct integration, and every model has a specific routing role (eco / balanced / sport, or specialty fallback). No middleman markup; no marketplace fees baked in.
Providers
All direct integrations. Excluded means registered but not currently routed to — e.g. DeepSeek is excluded until a customer asks for it, at which point we fund the account and flip it on.
| Provider | Architecture family | Models | Status |
|---|---|---|---|
| Anthropic | Claude | 5 | Active |
| Cerebras | Llama/Qwen | 2 | Active |
| DeepSeek | DeepSeek | 2 | Excluded |
| Fireworks | Kimi/GLM | 2 | Active |
| Gemini | 3 | Active | |
| Groq | Llama/Qwen/GPT-OSS | 5 | Active |
| Mistral | Mistral | 5 | Active |
| OpenAI | GPT | 3 | Active |
Models
Cost is per 1M tokens, before Prism markup. The "routes" column shows where each model is picked by mode-based auto-routing today.
| Model | Provider | Capability | Input $/1M | Output $/1M | Auto-router uses for |
|---|---|---|---|---|---|
| cerebras-llama-8b | Cerebras | Small (fast) | $0.10 | $0.10 | code/eco |
| cerebras-qwen-235b | Cerebras | Frontier | $0.60 | $1.20 | code/balanced |
| claude-haiku | Anthropic | Small (fast) | $0.80 | $4.00 | — |
| claude-opus | Anthropic | Frontier | $15.00 | $75.00 | simple/sport, reasoning/sport |
| claude-opus-4-7 | Anthropic | Frontier | $15.00 | $75.00 | — |
| claude-sonnet | Anthropic | Large | $3.00 | $15.00 | — |
| claude-sonnet-4-7 | Anthropic | Large | $3.00 | $15.00 | — |
| codestral | Mistral | Code-specialized | $0.30 | $0.90 | — |
| deepseek-v4-flash | DeepSeek | Small (fast) | $0.07 | $0.28 | — |
| deepseek-v4-pro | DeepSeek | Frontier | $0.14 | $0.56 | — |
| fireworks-glm-5p1 | Fireworks | Large | $0.50 | $2.00 | — |
| fireworks-kimi-k2 | Fireworks | Long-context | $0.60 | $2.50 | — |
| gemini-3-5-pro | Large | $1.25 | $10.00 | — | |
| gemini-flash | Small (fast) | $0.07 | $0.30 | — | |
| gemini-pro | Large | $1.25 | $10.00 | complex/sport | |
| gpt-4o | OpenAI | Large | $2.50 | $10.00 | complex/balanced |
| gpt-4o-mini | OpenAI | Small (fast) | $0.15 | $0.60 | — |
| gpt-5-5 | OpenAI | Frontier | $2.50 | $10.00 | — |
| groq-gpt-oss | Groq | Large | $0.15 | $0.75 | — |
| groq-llama-70b | Groq | Large | $0.59 | $0.79 | complex/eco |
| groq-llama-8b | Groq | Small (fast) | $0.05 | $0.08 | simple/eco, simple/balanced, reasoning/eco |
| groq-llama4-scout | Groq | Large | $0.11 | $0.34 | — |
| groq-qwen-32b | Groq | Medium | $0.29 | $0.59 | reasoning/balanced |
| magistral-medium | Mistral | Reasoning | $2.00 | $5.00 | — |
| ministral-8b | Mistral | Small (fast) | $0.10 | $0.10 | — |
| mistral-large | Mistral | Frontier | $2.00 | $6.00 | — |
| mistral-medium | Mistral | Medium | $0.40 | $2.00 | code/sport |
Try it
Specify a mode (eco / balanced / sport) and Prism picks the right model per request. Override with X-Prism-Model-Prefer to force a specific model (Pro+ for non-incumbent providers).