CoreWeave
CoreWeave serves 20 open models from 10 labs on OpenRouter, in 20 offers. Of the 20 models it shares with other providers, it has the lowest blended price on 3 (1 of them tied). Its median blended price is $0.2787 (₹26) per million tokens, blended 3:1 input to output. Its most common precision is 4-bit (10 of 20 offers).
Company and policies
- Headquarters listed by OpenRouter
- United States
- Datacenter countries listed by OpenRouter
- United States
Privacy policyTerms of serviceProvider on OpenRouter
Provider-level locations do not guarantee where a particular request is processed. Check the provider's terms for data handling. Directory source, checked 14 Nov 2026.
Precision and price rank
| Offers | 4-bit | 8-bit | 16-bit | Total |
|---|---|---|---|---|
| Offers | 10 | 6 | 4 | 20 |
| Rank | Models |
|---|---|
| Lowest or tied | 3 |
| 2nd | 1 |
| 3rd | 1 |
| 4th or lower | 15 |
Every model CoreWeave serves 20
USD and INR per million tokens| Precision | Lab | Pricing | ||||||
|---|---|---|---|---|---|---|---|---|
| $0.165₹16 | $0.13₹12 | $0.27₹25 | 8 of 23 | 262,144 | FP8 | DeepSeek | All providers | |
| $1.9725₹186 | $1.31₹124 | $3.96₹373 | 10 of 20 | 1,048,576 | FP8 | DeepSeek | All providers | |
| $0.26₹25 | $0.13₹12 | $0.65₹61 | 12 of 28 | 1,048,576 | FP8 | DeepSeek | All providers | |
| $0.525₹50 | $0.30₹28 | $1.20₹113 | 32 of 38 | 1,048,576 | FP8 | DeepSeek | All providers | |
| $0.15₹14 | $0.10₹9.43 | $0.30₹28 | 10 of 15 | 262,144 | BF16 | All providers | ||
| $0.16₹15 | $0.10₹9.43 | $0.34₹32 | 1 of 15Ties for lowest | 262,144 | FP4 | All providers | ||
| $1.07₹101 | $0.62₹58 | $2.42₹228 | 8 of 24 | 1,048,576 | FP4 | Z.ai (Zhipu) | All providers | |
| $1.3125₹124 | $0.75₹71 | $3.00₹283 | 16 of 37 | 1,048,576 | NVFP4 | Z.ai (Zhipu) | All providers | |
| $0.185₹17 | $0.12₹11 | $0.38₹36 | 15 of 37 | 1,048,576 | NVFP4 | Z.ai (Zhipu) | All providers | |
| $0.065₹6.13 | $0.03₹2.83 | $0.17₹16 | 4 of 22 | 131,072 | FP4 | OpenAI | All providers | |
| $0.055₹5.19 | $0.03₹2.83 | $0.13₹12 | 4 of 10 | 131,072 | FP4 | OpenAI | All providers | |
| $0.0825₹7.78 | $0.06₹5.66 | $0.15₹14 | 1 of 2Lowest | 131,072 | BF16 | IBM Granite | All providers | |
| $1.34₹126 | $0.65₹61 | $3.41₹322 | 5 of 15 | 262,144 | FP4 | Moonshot AI | All providers | |
| $1.20₹113 | $0.60₹57 | $3.00₹283 | 1 of 10Lowest | 262,144 | INT4 | Moonshot AI | All providers | |
| $0.71₹67 | $0.71₹67 | $0.71₹67 | 5 of 8 | 128,000 | FP16 | Meta | All providers | |
| $0.4125₹39 | $0.23₹22 | $0.96₹91 | 2 of 6 | 262,144 | FP4 | MiniMax | All providers | |
| $0.1025₹9.66 | $0.07₹6.60 | $0.20₹19 | 4 of 4 | 262,144 | BF16 | NVIDIA | All providers | |
| $0.23₹22 | $0.15₹14 | $0.47₹44 | 4 of 10 | 1,048,576 | NVFP4 | NVIDIA | All providers | |
| $0.2975₹28 | $0.10₹9.43 | $0.89₹84 | 3 of 10 | 262,144 | FP8 | Alibaba Qwen | All providers | |
| $1.05₹99 | $0.40₹38 | $3.00₹283 | 17 of 23 | 262,144 | FP8 | Alibaba Qwen | All providers |
Where CoreWeave is cheapest
3 of 20 models shared with other providers
gemma-4-31B-it$0.16 ₹15Ties for lowest of 15 providers
granite-4.2-8b$0.0825 ₹7.78Lowest of 2 providers
Kimi-K2.7-Code$1.20 ₹113Lowest of 10 providers
CoreWeave questions
Which open models does CoreWeave serve?
CoreWeave serves 20 open models from 10 labs on OpenRouter. The most downloaded on Hugging Face are Qwen3.8-27B, gemma-4-31B-it and GLM-5.3-Flash. Prices were fetched 14 Nov 2026.
Is CoreWeave the cheapest provider for any open model?
Yes, among models that other providers also serve. CoreWeave has the lowest blended price on 3 of 20: lowest on granite-4.2-8b and Kimi-K2.7-Code, and ties for the lowest on gemma-4-31B-it. For gemma-4-31B-it it charges $0.16 (₹15) per million tokens, blended 3:1 input to output.
What precision does CoreWeave serve models at?
Of its 20 offers, CoreWeave lists 4-bit for 10, 8-bit for 6 and 16-bit for 4. Prices are in US dollars; we show rupees at ₹94.29 per $1.
About this data
Every number on this page comes from the endpoints OpenRouter lists for each model, fetched 14 Nov 2026. INR is shown beside USD at ₹94.29 per $1. Prices change, so check the provider before you build on a number.