Models & pricing¶
How pricing works¶
- Prices are per million tokens, split into input / cached input / output.
- Cached input, prompt tokens served from cache, bills at that model's cached rate in the table. Details →
Rate card¶
The live rate card is in the portal model catalog and in the API,
/v1/models returns each model's verified context_length and an
pricing object (prompt / completion / input_cache_read, per-token USD strings):
curl https://api.keln.ai/v1/models
| Model | Context | Input $/M | Cached input $/M | Output $/M | Vision |
|---|---|---|---|---|---|
| deepseek-ai/deepseek-v4-flash-0731 | 1,048,576 | 0.14 | 0.07 | 0.28 | |
| deepseek-ai/deepseek-v4-pro-0813 | 1,048,576 | 1.32 | 0.12 | 3.96 | |
| google/gemma-4-31b-it | 1,048,576 | 0.15 | 0.12 | 0.46 | ✓ |
| minimaxai/minimax-m3 | 1,048,576 | 0.30 | 0.12 | 1.20 | ✓ |
| moonshotai/kimi-k2.7-code | 262,144 | 0.95 | 0.2375 | 4.00 | ✓ |
| moonshotai/kimi-k3 | 1,048,576 | 3.00 | 0.60 | 15.00 | ✓ |
| nvidia/nemotron-3-ultra-550b-a55b | 512,288 | 0.80 | 0.20 | 3.00 | |
| openai/gpt-oss-120b | 131,072 | 0.10 | 0.10 | 0.60 | |
| qwen/qwen3.6-35b-a3b | 262,144 | 0.25 | 0.20 | 1.49 | ✓ |
| qwen/qwen3.8-2.4t-a95b | 1,010,000 | 2.50 | 0.50 | 6.25 | |
| qwen/qwen3.8-27b | 1,000,000 | 0.50 | 0.06 | 3.20 | ✓ |
| stepfun-ai/step-3.7-flash | 262,144 | 0.20 | 0.05 | 1.15 | |
| xiaomimimo/mimo-v2.5 | 1,048,576 | 0.17 | 0.0425 | 0.33 | ✓ |
| xiaomimimo/mimo-v2.5-pro | 1,048,576 | 0.50 | 0.125 | 1.50 | |
| zai-org/glm-5.2 | 1,048,576 | 1.40 | 0.35 | 4.40 | |
| zai-org/glm-5.3-flash | 1,048,576 | 0.15 | 0.036 | 0.50 | ✓ |
This table mirrors the pricing block of /v1/models exactly, scrape the
endpoint rather than hardcoding these numbers if you'd like the updates for free. Ids are
canonically lowercase and matched case-insensitively.
Tools¶
| Tool | Price | Unit |
|---|---|---|
| Web search | $10.00 | per 1,000 searches, plus the tokens of the results the model reads |
Model lifecycle¶
Models are added when the supply pool can serve them at the published SLA. Deprecations get 14 days' notice by email to active users of that model.