Model catalogue
Prices per 1M tokens. Routing suffixes: :fast :cheap :eu work on any family.
| Model | Vendor | Context | Input | Output | p50 TTFT |
|---|---|---|---|---|---|
| lum-route:smart | Lumenal Router | — | $0.09 | $0.36 | 96 ms |
| lumen-core-70b | Lumenal · proprietary | 128k | $0.35 | $1.40 | 121 ms |
| lumen-core-8b | Lumenal · proprietary | 32k | $0.04 | $0.16 | 48 ms |
| llama-4.2-90b | Meta · open weights | 128k | $0.12 | $0.48 | 140 ms |
| qwen-turbo-next | Alibaba · open weights | 256k | $0.06 | $0.24 | 102 ms |
| deepseek-r2 | DeepSeek · open weights | 128k | $0.27 | $1.10 | 180 ms |
| gpt-5 | OpenAI | 400k | $1.25 | $10.00 | 210 ms |
| gpt-5-mini | OpenAI | 400k | $0.25 | $2.00 | 130 ms |
| claude-sonnet-4.5 | Anthropic | 200k | $3.00 | $15.00 | 240 ms |
| gemini-3-flash | 1M | $0.10 | $0.40 | 88 ms | |
| devstral-small-2 | Mistral · open weights | 128k | $0.08 | $0.30 | 95 ms |
| mistral-embed-v3 | Mistral | 8k | $0.02 | — | 31 ms |
38 models total · open-weight entries include checksummed weight mirrors (24 GB VRAM minimum in 4-bit). Embeddings, moderation and image endpoints listed below the fold.