# DeepInfra API pricing and models

DeepInfra publishes per-token prices for 92 text models, as of October 20, 2026. The lowest output rate is $0.03 per million tokens, on Mistral-Nemo-Instruct-2407, and the highest-scoring model it serves is Claude Opus 5, at 88 on our 100-point score.

DeepInfra runs a serverless, OpenAI-compatible API for open-weight models and bills per token.

Source: [DeepInfra pricing page](https://deepinfra.com/pricing), read 2026-10-20. Prices in US dollars per million tokens.

## Models we score

| Model | Score | Input | Output | Cached input | Context | vs maker |
|---|---|---|---|---|---|---|
| claude-opus-5 (`anthropic/claude-opus-5`) | 88 | $5.00 | $25.00 | — | 1M | Same |
| claude-sonnet-5-5 (`anthropic/claude-sonnet-5-5`) | 83 | $2.00 | $10.00 | — | 1M | Same |
| claude-fable-5 (`anthropic/claude-fable-5`) | 83 | $10.00 | $50.00 | — | 1M | Same |
| claude-opus-5-5 (`anthropic/claude-opus-5-5`) | 82 | $4.00 | $20.00 | $0.20 | 1M | Same |
| Qwen3.8-Max (`Qwen/Qwen3.8-Max`) | 81 | $1.65 | $4.95 | $0.21 | 256K | — |
| Kimi-K3 (`moonshotai/Kimi-K3`) | 79 | $2.85 | $14.25 | $0.28 | 1.05M | -5% |
| Step-5-Preview (`stepfun-ai/Step-5-Preview`) | 77 | $1.00 | $2.70 | $0.10 | 1.05M | Same |
| gemini-3.7-flash (`google/gemini-3.7-flash`) | 77 | $0.75 | $3.75 | — | 1M | Same |
| claude-opus-4-8 (`anthropic/claude-opus-4-8`) | 77 | $5.00 | $25.00 | — | 1M | Same |
| Qwen3.7-Max (`Qwen/Qwen3.7-Max`) | 76 | $2.50 | $7.50 | $0.50 | 256K | — |
| claude-sonnet-5 (`anthropic/claude-sonnet-5`) | 76 | $3.00 | $15.00 | — | 1M | +50% |
| Ling-3.1-flash (`inclusionAI/Ling-3.1-flash`) | 74 | $0.075 | $0.23 | $0.015 | 262K | — |
| DeepSeek-V4.1-Flash-1017 (`deepseek-ai/DeepSeek-V4.1-Flash-1017`) | 73 | $0.098 | $0.29 | $0.003 | 1.05M | -75% |
| DeepSeek-V4.1-Flash (`deepseek-ai/DeepSeek-V4.1-Flash`) | 73 | $0.14 | $0.42 | $0.004 | 1.05M | -65% |
| MiMo-V2.6-Pro (`XiaomiMiMo/MiMo-V2.6-Pro`) | 73 | $0.39 | $0.78 | $0.003 | 1.05M | -10% |
| Qwen3.8-27B (`Qwen/Qwen3.8-27B`) | 73 | $0.15 | $1.88 | $0.037 | 262K | — |
| Hy4-preview (`tencent/Hy4-preview`) | 72 | $0.83 | $2.50 | $0.042 | 1.05M | — |
| gemini-3.1-pro (`google/gemini-3.1-pro`) | 71 | $2.00 | $12.00 | — | 1M | Same |
| gemini-3.5-flash (`google/gemini-3.5-flash`) | 70 | $1.50 | $9.00 | — | 1M | Same |
| MiMo-V2.6-Flash (`XiaomiMiMo/MiMo-V2.6-Flash`) | 69 | $0.13 | $0.27 | $0.003 | 1.05M | -5% |
| GLM-5.3-Flash (`zai-org/GLM-5.3-Flash`) | 69 | $0.15 | $0.50 | $0.03 | 1.05M | — |
| DeepSeek-V4-Pro-0813 (`deepseek-ai/DeepSeek-V4-Pro-0813`) | 69 | $1.30 | $2.60 | $0.10 | 1.05M | -34% |
| GLM-5.3 (`zai-org/GLM-5.3`) | 68 | $0.56 | $2.50 | $0.13 | 1.05M | — |
| Kimi-K2.6 (`moonshotai/Kimi-K2.6`) | 67 | $0.75 | $3.50 | $0.15 | 262K | -12% |
| Ling-3.0-flash-VL (`inclusionAI/Ling-3.0-flash-VL`) | 66 | $0.06 | $0.18 | $0.012 | 131K | — |
| MiniMax-M3 (`MiniMaxAI/MiniMax-M3`) | 65 | $0.28 | $1.10 | $0.056 | 524K | -8% |
| GLM-5.2 (`zai-org/GLM-5.2`) | 65 | $0.56 | $1.80 | $0.10 | 1.05M | -59% |
| Inkling-Small (`thinkingmachines/Inkling-Small`) | 64 | $0.45 | $1.20 | $0.10 | 524K | -17% |
| claude-sonnet-4-6 (`anthropic/claude-sonnet-4-6`) | 63 | $3.00 | $15.00 | — | 1M | Same |
| DeepSeek-V4-Flash-0731 (`deepseek-ai/DeepSeek-V4-Flash-0731`) | 62 | $0.06 | $0.18 | $0.015 | 1.05M | -36% |
| Qwen3.6-27B (`Qwen/Qwen3.6-27B`) | 62 | $0.32 | $3.20 | — | 262K | — |
| Llama-3.3-70B-Instruct-Turbo (`meta-llama/Llama-3.3-70B-Instruct-Turbo`) | 61 | $0.10 | $0.32 | — | 131K | — |
| DeepSeek-V3.2 (`deepseek-ai/DeepSeek-V3.2`) | 61 | $0.26 | $0.38 | $0.13 | — | -10% |
| Muse-Glimmer-30B (`meta-models/Muse-Glimmer-30B`) | 61 | $0.30 | $1.20 | $0.04 | 131K | — |
| Ling-3.0-flash (`inclusionAI/Ling-3.0-flash`) | 60 | $0.06 | $0.18 | $0.012 | 131K | — |
| DeepSeek-R1-0528 (`deepseek-ai/DeepSeek-R1-0528`) | 60 | $0.50 | $2.15 | $0.35 | 164K | — |
| Qwen3.5-27B (`Qwen/Qwen3.5-27B`) | 59 | $0.26 | $2.60 | — | 262K | — |
| gemma-4-26B-A4B-it (`google/gemma-4-26B-A4B-it`) | 58 | $0.07 | $0.34 | — | 262K | — |
| gemini-2.5-pro (`google/gemini-2.5-pro`) | 58 | $1.25 | $10.00 | — | 1M | Same |
| Qwen3-32B (`Qwen/Qwen3-32B`) | 57 | $0.08 | $0.28 | — | 41K | — |
| gemini-3.1-flash-lite (`google/gemini-3.1-flash-lite`) | 57 | $0.25 | $1.50 | — | 1M | Same |
| gemini-2.5-flash (`google/gemini-2.5-flash`) | 57 | $0.30 | $2.50 | — | 1M | Same |
| claude-opus-4-7 (`anthropic/claude-opus-4-7`) | 57 | $5.00 | $25.00 | — | 1M | Same |
| gemma-4-31B-it-turbo (`google/gemma-4-31B-it-turbo`) | 56 | $0.09 | $0.34 | $0.05 | 262K | — |
| gemma-4-31B-it (`google/gemma-4-31B-it`) | 56 | $0.20 | $0.40 | — | 262K | — |
| gpt-oss-120b (`openai/gpt-oss-120b`) | 55 | $0.037 | $0.17 | — | 131K | — |
| Qwen3.6-35B-A3B (`Qwen/Qwen3.6-35B-A3B`) | 55 | $0.10 | $0.95 | $0.10 | 262K | — |
| Qwen3.5-35B-A3B (`Qwen/Qwen3.5-35B-A3B`) | 55 | $0.14 | $1.00 | $0.05 | 262K | — |
| Qwen3-Coder-480B-A35B-Instruct-Turbo (`Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo`) | 54 | $0.30 | $1.00 | $0.10 | 262K | — |
| DeepSeek-V3.1 (`deepseek-ai/DeepSeek-V3.1`) | 53 | $0.25 | $0.95 | $0.13 | 164K | — |
| GLM-4.7 (`zai-org/GLM-4.7`) | 53 | $0.40 | $1.75 | $0.08 | 203K | — |
| GLM-4.6 (`zai-org/GLM-4.6`) | 53 | $0.50 | $2.00 | $0.10 | 203K | — |
| gpt-oss-20b (`openai/gpt-oss-20b`) | 50 | $0.03 | $0.14 | — | 131K | — |
| granite-4.2-8b (`ibm-granite/granite-4.2-8b`) | 50 | $0.06 | $0.25 | $0.015 | 131K | — |
| claude-haiku-4-5 (`anthropic/claude-haiku-4-5`) | 47 | $1.00 | $5.00 | — | 200K | Same |
| DeepSeek-V3 (`deepseek-ai/DeepSeek-V3`) | 43 | $0.32 | $0.89 | — | 164K | -19% |
| gemma-3-27b-it (`google/gemma-3-27b-it`) | 41 | $0.08 | $0.16 | — | 131K | — |
| phi-4 (`microsoft/phi-4`) | 37 | $0.07 | $0.14 | — | 16K | — |
| Qwen2.5-72B-Instruct (`Qwen/Qwen2.5-72B-Instruct`) | 0 | $0.36 | $0.40 | — | 33K | — |
| Seed-2.0-mini (`ByteDance/Seed-2.0-mini`) | 0 | $0.10 | $0.40 | $0.02 | 256K | — |

## Other models

| Model | Input | Output | Cached input | Context |
|---|---|---|---|---|
| Mistral-Nemo-Instruct-2407 (`mistralai/Mistral-Nemo-Instruct-2407`) | $0.029 | $0.03 | — | 131K |
| Meta-Llama-3.1-8B-Instruct-Turbo (`meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo`) | $0.02 | $0.04 | — | 131K |
| L3-8B-Lunaris-v1-Turbo (`Sao10K/L3-8B-Lunaris-v1-Turbo`) | $0.04 | $0.05 | — | 8K |
| Mistral-Small-24B-Instruct-2501 (`mistralai/Mistral-Small-24B-Instruct-2501`) | $0.05 | $0.08 | — | 33K |
| gemma-3-4b-it (`google/gemma-3-4b-it`) | $0.05 | $0.10 | — | 131K |
| gemma-3-12b-it (`google/gemma-3-12b-it`) | $0.05 | $0.15 | — | 131K |
| Qwen3.5-9B (`Qwen/Qwen3.5-9B`) | $0.10 | $0.15 | — | 262K |
| NVIDIA-Nemotron-3.5-Lightning (`nvidia/NVIDIA-Nemotron-3.5-Lightning`) | $0.06 | $0.16 | $0.03 | 262K |
| DeepSeek-V4-Flash (`deepseek-ai/DeepSeek-V4-Flash`) | $0.09 | $0.18 | $0.018 | 1.05M |
| Mistral-Small-3.2-24B-Instruct-2506 (`mistralai/Mistral-Small-3.2-24B-Instruct-2506`) | $0.075 | $0.20 | — | 128K |
| Qwen3-14B (`Qwen/Qwen3-14B`) | $0.12 | $0.24 | — | 41K |
| Llama-4-Scout-17B-16E-Instruct (`meta-llama/Llama-4-Scout-17B-16E-Instruct`) | $0.10 | $0.30 | — | 328K |
| Qwen3.8-Flash (`Qwen/Qwen3.8-Flash`) | $0.11 | $0.38 | $0.014 | 1M |
| Meta-Llama-3.1-70B-Instruct-Turbo (`meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo`) | $0.40 | $0.40 | — | 131K |
| MythoMax-L2-13b (`Gryphe/MythoMax-L2-13b`) | $0.40 | $0.40 | — | 4K |
| NVIDIA-Nemotron-3-Super-120B-A12B (`nvidia/NVIDIA-Nemotron-3-Super-120B-A12B`) | $0.085 | $0.40 | — | 262K |
| NVIDIA-Nemotron-3.5-Super-VL (`nvidia/NVIDIA-Nemotron-3.5-Super-VL`) | $0.15 | $0.50 | $0.02 | 262K |
| Qwen3-30B-A3B (`Qwen/Qwen3-30B-A3B`) | $0.12 | $0.50 | — | 41K |
| Qwen3-235B-A22B-Instruct-2507 (`Qwen/Qwen3-235B-A22B-Instruct-2507`) | $0.09 | $0.55 | — | 262K |
| Qwen3-VL-30B-A3B-Instruct (`Qwen/Qwen3-VL-30B-A3B-Instruct`) | $0.15 | $0.60 | — | 262K |
| DeepSeek-V4-Flash-Vision-Exp (`deepseek-ai/DeepSeek-V4-Flash-Vision-Exp`) | $0.22 | $0.65 | $0.007 | 1.05M |
| gemma-4-31B-it-Ultra (`google/gemma-4-31B-it-Ultra`) | $0.27 | $0.76 | — | 131K |
| L3.1-70B-Euryale-v2.2 (`Sao10K/L3.1-70B-Euryale-v2.2`) | $0.85 | $0.85 | — | 131K |
| Qwen3-VL-235B-A22B-Instruct (`Qwen/Qwen3-VL-235B-A22B-Instruct`) | $0.20 | $0.88 | $0.11 | 262K |
| gpt-oss-120b-Ultra (`openai/gpt-oss-120b-Ultra`) | $0.20 | $0.95 | — | 131K |
| Qwen3-Next-80B-A3B-Instruct (`Qwen/Qwen3-Next-80B-A3B-Instruct`) | $0.09 | $1.10 | — | 262K |
| Seed-1.8 (`ByteDance/Seed-1.8`) | $0.25 | $2.00 | $0.05 | 256K |
| NVIDIA-Nemotron-3-Ultra-550B-A55B (`nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B`) | $0.50 | $2.20 | $0.10 | 262K |
| DeepSeek-V4-Pro (`deepseek-ai/DeepSeek-V4-Pro`) | $1.30 | $2.60 | $0.10 | 1.05M |
| Qwen3.5-397B-A17B (`Qwen/Qwen3.5-397B-A17B`) | $0.45 | $3.00 | $0.22 | 262K |
| Seed-2.0-code (`ByteDance/Seed-2.0-code`) | $0.50 | $3.00 | $0.10 | 256K |
| Seed-2.0-pro (`ByteDance/Seed-2.0-pro`) | $0.50 | $3.00 | $0.10 | 256K |

## What we measured (2026-09-01)

DeepInfra billed the lowest cost per GLM-5.2 token in the run, $0.38 per million blended, because it cached 53% of our Flash input and about half of our GLM input at roughly 80% off with no caching parameter in the request. Median latency was 2.4 to 5.3 seconds, the slowest of the five scored platforms on GLM-5.2. All 900 requests succeeded.

| Model run | Invoiced /1M blended | Median latency | p95 latency | Succeeded |
|---|---|---|---|---|
| DeepSeek V4 Flash | $0.058 | 2.42 s | 6.93 s | 100.0% |
| GLM-5.2 | $0.377 | 5.29 s | 13.82 s | 100.0% |

Method and invoices: https://benchlm.ai/openrouter-alternatives

## Hosts that sell the same models

- [OpenRouter](https://benchlm.ai/hosts/openrouter): 57 shared, 15 cheaper per output token (GLM-5.3-Flash: $0.12 vs $0.50)
- [Novita AI](https://benchlm.ai/hosts/novita): 32 shared, 3 cheaper per output token (Qwen3.7 Max: $3.75 vs $7.50)
- [Together AI](https://benchlm.ai/hosts/together): 13 shared, 0 cheaper per output token
- [Nebius Token Factory](https://benchlm.ai/hosts/nebius): 12 shared, 0 cheaper per output token
- [Baseten](https://benchlm.ai/hosts/baseten): 8 shared, 0 cheaper per output token
- [Fireworks AI](https://benchlm.ai/hosts/fireworks): 7 shared, 0 cheaper per output token

## FAQ

### How much does the DeepInfra API cost?

DeepInfra bills per million tokens, with separate input and output rates for each model. On October 20, 2026 its output rates ran from $0.03 on Mistral-Nemo-Instruct-2407 to $50.00 across 92 paid models. Input is usually cheaper than output, so a long prompt with a short reply costs less than the output rate suggests.

### Which models can you run on DeepInfra?

92 text models are listed, and 58 of them match a model we score. The highest-scoring is Claude Opus 5 at 88. The table on this page has every model with its input, output and cached-input rate, the context window, and the date we read the price.

### Is DeepInfra fast?

In our paid September 1, 2026 run, DeepInfra returned a median of 2.42 seconds on DeepSeek V4 Flash and 5.29 seconds on GLM-5.2, over 450 requests per model. That is one workload on one date, so check your own prompts before you rely on it.

### What are the alternatives to DeepInfra?

OpenRouter lists 57 of the same models and charges less output per token on 15 of them. Novita AI and Together AI also overlap. Price is only part of the choice: speed, rate limits and which model snapshot a host serves differ too.

Canonical page: https://benchlm.ai/hosts/deepinfra