# DeepInfra vs Fireworks AI

Canonical: https://inetgeek.com/compare/deepinfra-vs-fireworks/

Every value below is read from DeepInfra's and Fireworks AI's own documentation. See https://inetgeek.com/methodology/ for how.

## At a glance

| Criterion | DeepInfra | Fireworks AI |
| --- | --- | --- |
| Pricing model | Two products on one bill: per-token serverless inference with a separate cached-input rate for each model, and dedicated GPUs billed per minute and invoiced weekly. Models without per-token pricing are billed for inference execution time instead. | Three products on one bill. Serverless inference is per token, per model. Fine-tuning is per million TRAINING tokens by parameter band — $0.50 for models up to 16B, $10.00 above 300B. Dedicated GPUs are hourly: $7.00 for an H100 or H200, $10.00 for a B200. |
| Input price, top model | $1.30 per million input tokens for DeepSeek-V4-Pro, the most expensive model in its catalogue. Cached input is $0.10. | $3.00 per million input tokens for Kimi K3, the most expensive model in its serverless catalogue. Fireworks hosts other labs' open-weight models rather than its own, so "flagship" here means the top of what it serves. |
| Output price, top model | $2.60 per million output tokens for DeepSeek-V4-Pro. | $15.00 per million output tokens for Kimi K3. |
| Input price, cheapest model | $0.06 per million input tokens for DeepSeek-V4-Flash-0731, with cached input at $0.015 and output at $0.18. DeepSeek itself charges $0.44 at peak for the same checkpoint and Fireworks $0.22. | $0.15 per million input tokens for GLM 5.3 Flash, with output at $0.50. DeepSeek-V4-Flash is $0.22 in and $0.66 out — the same model DeepSeek itself sells at $0.44 peak. |
| Context window | 1024k tokens on the DeepSeek V4 models it serves; 160k on the V3 generation. The window is the model's rather than a platform limit. | 1,048,576 tokens on Kimi K3, GLM 5.3 and both DeepSeek V4 variants; 262,144 on Qwen3.8 and 131,072 on Muse Glimmer 30B. The window is the model's, not Fireworks'. |
| Prompt caching | Supported | Supported |
| H100 SXM, per GPU-hour | $2.20 per GPU-hour for a dedicated H100 80GB — below every dedicated cloud in this dataset. A100 80GB is $0.89, H200 $2.69, B200 $3.69. Billed in minute granularity, invoiced weekly. | $7.00 per hour for an H100 80GB on demand, or $8.00 on the shorter commitment shown beside it. The same rate applies to an H200 141GB. |
| Largest GPU offered | 270GB per GPU on a B300, at $4.89 per GPU-hour. | 180GB per GPU on a B200, at $10.00 an hour on demand. |

## Where they differ

### Pricing model

- DeepInfra: Two products on one bill: per-token serverless inference with a separate cached-input rate for each model, and dedicated GPUs billed per minute and invoiced weekly. Models without per-token pricing are billed for inference execution time instead. ([source](https://deepinfra.com/pricing))
- Fireworks AI: Three products on one bill. Serverless inference is per token, per model. Fine-tuning is per million TRAINING tokens by parameter band — $0.50 for models up to 16B, $10.00 above 300B. Dedicated GPUs are hourly: $7.00 for an H100 or H200, $10.00 for a B200. ([source](https://fireworks.ai/pricing))

### Input price, top model

- DeepInfra: $1.30 per million input tokens for DeepSeek-V4-Pro, the most expensive model in its catalogue. Cached input is $0.10. ([source](https://deepinfra.com/pricing))
- Fireworks AI: $3.00 per million input tokens for Kimi K3, the most expensive model in its serverless catalogue. Fireworks hosts other labs' open-weight models rather than its own, so "flagship" here means the top of what it serves. ([source](https://fireworks.ai/models))

### Output price, top model

- DeepInfra: $2.60 per million output tokens for DeepSeek-V4-Pro. ([source](https://deepinfra.com/pricing))
- Fireworks AI: $15.00 per million output tokens for Kimi K3. ([source](https://fireworks.ai/models))

### Input price, cheapest model

- DeepInfra: $0.06 per million input tokens for DeepSeek-V4-Flash-0731, with cached input at $0.015 and output at $0.18. DeepSeek itself charges $0.44 at peak for the same checkpoint and Fireworks $0.22. ([source](https://deepinfra.com/pricing))
- Fireworks AI: $0.15 per million input tokens for GLM 5.3 Flash, with output at $0.50. DeepSeek-V4-Flash is $0.22 in and $0.66 out — the same model DeepSeek itself sells at $0.44 peak. ([source](https://fireworks.ai/models))

### Context window

- DeepInfra: 1024k tokens on the DeepSeek V4 models it serves; 160k on the V3 generation. The window is the model's rather than a platform limit. ([source](https://deepinfra.com/pricing))
- Fireworks AI: 1,048,576 tokens on Kimi K3, GLM 5.3 and both DeepSeek V4 variants; 262,144 on Qwen3.8 and 131,072 on Muse Glimmer 30B. The window is the model's, not Fireworks'. ([source](https://fireworks.ai/models))

### H100 SXM, per GPU-hour

- DeepInfra: $2.20 per GPU-hour for a dedicated H100 80GB — below every dedicated cloud in this dataset. A100 80GB is $0.89, H200 $2.69, B200 $3.69. Billed in minute granularity, invoiced weekly. ([source](https://deepinfra.com/pricing))
- Fireworks AI: $7.00 per hour for an H100 80GB on demand, or $8.00 on the shorter commitment shown beside it. The same rate applies to an H200 141GB. ([source](https://fireworks.ai/pricing))

### Largest GPU offered

- DeepInfra: 270GB per GPU on a B300, at $4.89 per GPU-hour. ([source](https://deepinfra.com/pricing))
- Fireworks AI: 180GB per GPU on a B200, at $10.00 an hour on demand. ([source](https://fireworks.ai/pricing))

## Which should you choose?

Pick DeepInfra if Running open-weight models at the lowest published per-token rates here — DeepSeek-V4-Flash at $0.06 per million input is a fraction of what the lab itself charges — with dedicated H100s at $2.20 an hour if throughput demands it.

Pick Fireworks AI if Running open-weight models without operating GPUs — the same DeepSeek and Qwen checkpoints the labs publish, served per token, with a path to dedicated hardware if throughput demands it.

Consider something else: DeepInfra — Only some models carry per-token pricing; the rest bill for inference execution time, which is a different and harder thing to forecast.

Consider something else: Fireworks AI — You want a frontier proprietary model: Fireworks serves other people's open weights, so the ceiling is whatever the open ecosystem has released.

## Documented by only one

- Free tier: Fireworks AI: Yes
- Max output tokens: Fireworks AI: No separate output cap. max_tokens is bounded by the model's context length — prompt plus completion must fit inside it — and what happens on overflow is set by context_length_exceeded_behavior rather than a fixed ceiling.

## Questions this comparison answers

**Should I choose DeepInfra or Fireworks AI?**

Pick DeepInfra if Running open-weight models at the lowest published per-token rates here — DeepSeek-V4-Flash at $0.06 per million input is a fraction of what the lab itself charges — with dedicated H100s at $2.20 an hour if throughput demands it.
Pick Fireworks AI if Running open-weight models without operating GPUs — the same DeepSeek and Qwen checkpoints the labs publish, served per token, with a path to dedicated hardware if throughput demands it.