Parasail: Models and Prices
Inference and training platform that orchestrates GPU capacity from data center partners.
Parasail served 2.67T tokens of other labs' open-weight models over Oct 3 to Oct 9, 2026 through a large public LLM router, #6 among providers.
Open-weight models
| Model | Family | Share of model | Tokens, 7d | Input $/M | Output $/M | Tokens/s | Uptime 3d | Context |
|---|---|---|---|---|---|---|---|---|
| GLM 5.3 Flash | GLM | 10.3% | 1.15T | $0.15 | $0.50 | 95 | 99.9% | 1.05M |
| DeepSeek V4.1 Flash | DeepSeek | 2.9% | 1.10T | $0.30 | $1.20 | 188 | 100.0% | 1.05M |
| Kimi K3 | Kimi | 6.3% | 122B | $2.60 | $13.00 | 64 | 99.5% | 1.05M |
| DeepSeek V4 Flash 0423 | DeepSeek | 2.7% | 82B | $0.14 | $0.28 | 63 | 100.0% | 1.05M |
| DeepSeek V4 Flash 0731 | DeepSeek | 1.5% | 78B | $0.14 | $0.28 | 94 | 100.0% | 1.05M |
| Gemma 3 27B | Gemma | 81.8% | 36B | $0.08 | $0.45 | 40 | 99.7% | 131K |
| gpt-oss-20b | gpt-oss | 14.7% | 30B | $0.03 | $0.15 | 65 | 99.9% | 131K |
| Gemma 4 31B | Gemma | 6.3% | 27B | $0.15 | $0.40 | 24 | 99.9% | 262K |
| Qwen3.8 27B | Qwen | 3.1% | 18B | $0.24 | $2.20 | 61 | 100.0% | 262K |
| Qwen3 Coder Next | Qwen | 100.0% | 11B | $0.12 | $0.80 | 40 | 100.0% | 262K |
| DeepSeek V4 Pro 0423 | DeepSeek | 1.5% | 9B | $0.45 | $3.48 | 38 | 94.1% | 1.05M |
| Qwen3.6 35B A3B | Qwen | 11.6% | 5B | $0.15 | $1.00 | 76 | 99.9% | 262K |
| GLM 5.3 | GLM | - | - | $1.40 | $4.40 | 74 | 99.7% | 1.05M |
| GLM 5.2 | GLM | - | - | $1.40 | $4.40 | 102 | 99.8% | 262K |
| MiniMax M3 | MiniMax | - | - | $0.30 | $1.20 | 74 | 99.9% | 1.05M |
| DeepSeek V4 Pro 0813 | DeepSeek | - | - | $1.32 | $3.96 | 68 | 100.0% | 1.05M |
| gpt-oss-120b | gpt-oss | - | - | $0.10 | $0.75 | 129 | 100.0% | 131K |
| Gemma 4 26B A4B | Gemma | - | - | $0.13 | $0.40 | 59 | 99.6% | 262K |
| Mistral Nemo | Mistral | - | - | $0.03 | $0.03 | 103 | 100.0% | 131K |
| Kimi K2.6 | Kimi | - | - | $0.75 | $3.50 | 59 | 100.0% | 262K |
| Qwen3 235B A22B Instruct 2507 | Qwen | - | - | $0.14 | $0.80 | 37 | 100.0% | 131K |
| Llama 3.3 70B Instruct | Llama | - | - | $0.22 | $0.50 | 50 | 100.0% | 131K |
| Qwen3.5-9B | Qwen | - | - | $0.10 | $0.25 | 114 | 100.0% | 262K |
| Qwen3.5 397B A17B | Qwen | - | - | $0.50 | $3.60 | 66 | 99.7% | 262K |
| Mistral Small 3.2 24B | Mistral | - | - | $0.09 | $0.30 | 50 | 100.0% | 131K |
| Qwen3 Next 80B A3B Instruct | Qwen | - | - | $0.10 | $1.10 | 56 | 100.0% | 262K |
| Qwen3 VL 235B A22B Instruct | Qwen | - | - | $0.21 | $1.90 | 30 | 99.5% | 131K |
| Llama 4 Maverick | Llama | - | - | $0.35 | $1.00 | 55 | 100.0% | 524K |
| Qwen3.5-35B-A3B | Qwen | - | - | $0.15 | $1.00 | 180 | 100.0% | 262K |
| Qwen3 VL 8B Instruct | Qwen | - | - | $0.25 | $0.75 | 44 | 87.0% | 262K |
| Llama 3 8B Lunaris | Other open-weight | - | - | $0.04 | $0.05 | 98 | 99.9% | 8K |
| Qwen2.5 VL 72B Instruct | Qwen | - | - | $0.80 | $1.00 | 40 | 98.8% | 128K |
| Cydonia 24B V4.1 | Other open-weight | - | - | $0.30 | $0.50 | 50 | 99.9% | 131K |
| Llama 3.2 3B Instruct | Llama | - | - | $0.05 | $0.33 | 263 | 99.9% | 131K |
| MythoMax 13B | Other open-weight | - | - | $0.08 | $0.11 | 98 | 100.0% | 4K |
| UnslopNemo 12B | Other open-weight | - | - | $0.40 | $0.40 | 59 | 99.8% | 1.02M |
| Skyfall 36B V2 | Other open-weight | - | - | $0.55 | $0.80 | 68 | 100.0% | 33K |
| UI-TARS 7B | Other open-weight | - | - | $0.10 | $0.20 | 62 | 99.9% | 128K |
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
- parasail.io
- Headquarters
- United States
- Latest financing
- $32M Series A co-led by Touring Capital and Kindred Ventures (April 2026); about $42M raised in total.
Sources: DCD.
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