CoreWeave: Models and Prices
GPU cloud (Nasdaq: CRWV) that also serves open-weight models as a token API.
CoreWeave served 1.55T tokens of other labs' open-weight models over Oct 3 to Oct 9, 2026 through a large public LLM router, #11 among providers.
Open-weight models
| Model | Family | Share of model | Tokens, 7d | Input $/M | Output $/M | Tokens/s | Uptime 3d | Context |
|---|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash | DeepSeek | 1.2% | 451B | $0.20 | $0.65 | 179 | 99.5% | 1.05M |
| DeepSeek V4 Flash 0731 | DeepSeek | 8.3% | 422B | $0.13 | $0.28 | 108 | 100.0% | 262K |
| GLM 5.3 Flash | GLM | 2.5% | 277B | $0.15 | $0.50 | 137 | 99.7% | 1.05M |
| Gemma 4 31B | Gemma | 22.4% | 95B | $0.10 | $0.34 | 68 | 100.0% | 262K |
| DeepSeek V4 Pro 0813 | DeepSeek | 6.3% | 67B | $1.31 | $3.96 | 110 | 99.9% | 1.05M |
| gpt-oss-120b | gpt-oss | 15.0% | 63B | $0.03 | $0.17 | 42 | 99.8% | 131K |
| gpt-oss-20b | gpt-oss | 29.3% | 60B | $0.03 | $0.13 | 128 | 100.0% | 131K |
| MiniMax M3 | MiniMax | 4.6% | 55B | $0.23 | $0.96 | 68 | 99.9% | 262K |
| GLM 5.2 | GLM | 3.4% | 52B | $0.76 | $2.42 | 161 | 99.9% | 1.05M |
| Kimi K2.6 | Kimi | 7.2% | 7B | $0.65 | $3.41 | 64 | 99.8% | 262K |
| Gemma 4 26B A4B | Gemma | 0.6% | 2B | $0.10 | $0.30 | 63 | 99.8% | 262K |
| Nemotron 3.5 Lightning | Nemotron | - | - | $0.07 | $0.20 | 235 | 100.0% | 262K |
| Qwen3.8 27B | Qwen | - | - | $0.40 | $3.00 | 110 | 98.8% | 262K |
| Qwen3.6 35B A3B | Qwen | - | - | $0.25 | $1.25 | 136 | 99.9% | 262K |
| DeepSeek V3.1 | DeepSeek | - | - | $0.55 | $1.65 | 57 | 100.0% | 161K |
| Kimi K2.7 Code | Kimi | - | - | $0.71 | $3.50 | 27 | 100.0% | 262K |
| Llama 3.1 8B Instruct | Llama | - | - | $0.22 | $0.22 | 143 | 99.9% | 131K |
| Llama 3.3 70B Instruct | Llama | - | - | $0.71 | $0.71 | 63 | 99.7% | 128K |
| Granite 4.2 8B | Other open-weight | - | - | $0.10 | $0.15 | 82 | 100.0% | 131K |
List prices per million tokens as of October 9, 2026. Share: of all tokens served on that model; a dash where the volume is not reported.
Company
- Website
- coreweave.com
- Headquarters
- Livingston, NJ
- Compute
- Operates its own GPU fleet across leased and owned data centers.
Sources: CoreWeave.
Finance your GPUs
Serving tokens on your own GPUs? Tell us the cluster, the models you serve and the contracts behind them, and we will share it with partner funders that fit.
Prefer email? hello@amcompute.com