# Fireworks AI vs Novita AI

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

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

## At a glance

| Criterion | Fireworks AI | Novita AI |
| --- | --- | --- |
| Pricing model | 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. | Per-token serverless inference with a cache-read rate published beside most models, and GPU instances offered at both on-demand and spot rates — the spot column is roughly half. Two models, Ling 3.0 Flash Fin and Sante, are listed at no charge on both input and output. |
| Input price, top model | $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. | $2.00 per million input tokens for Qwen3.8 Max, the most expensive model in its catalogue, with cache reads at $0.25 and output at $6.00. |
| Output price, top model | $15.00 per million output tokens for Kimi K3. | $6.00 per million output tokens for Qwen3.8 Max. |
| Input price, cheapest model | $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. | $0.075 per million input tokens for GLM 5.3 Flash at short context, rising to $0.15 above the threshold, with cache reads at $0.015 and output at $0.25. Fireworks serves the same model at $0.15. |
| Context window | 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'. | 1M tokens on the DeepSeek V4 line and GLM 5.3, published as "1M" rather than an exact figure — so it is recorded at 1,000,000 and may in practice be the 1,048,576 other providers state. Qwen3.8 is listed at 977K. |
| Prompt caching | Supported | Supported |
| H100 SXM, per GPU-hour | $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. | $3.39 per GPU-hour for an on-demand H100 SXM 80GB, or $1.70 spot. The page names the card the site's GPU comparison is fixed on, so no substitution is involved. |
| Largest GPU offered | 180GB per GPU on a B200, at $10.00 an hour on demand. | 80GB per GPU on an H100 SXM, the largest card it lists; below it are an L40S 48GB, RTX 5090 32GB and RTX 4090 24GB. |

## Where they differ

### Pricing model

- 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))
- Novita AI: Per-token serverless inference with a cache-read rate published beside most models, and GPU instances offered at both on-demand and spot rates — the spot column is roughly half. Two models, Ling 3.0 Flash Fin and Sante, are listed at no charge on both input and output. ([source](https://novita.ai/pricing))

### Input price, top model

- 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))
- Novita AI: $2.00 per million input tokens for Qwen3.8 Max, the most expensive model in its catalogue, with cache reads at $0.25 and output at $6.00. ([source](https://novita.ai/pricing))

### Output price, top model

- Fireworks AI: $15.00 per million output tokens for Kimi K3. ([source](https://fireworks.ai/models))
- Novita AI: $6.00 per million output tokens for Qwen3.8 Max. ([source](https://novita.ai/pricing))

### Input price, cheapest model

- 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))
- Novita AI: $0.075 per million input tokens for GLM 5.3 Flash at short context, rising to $0.15 above the threshold, with cache reads at $0.015 and output at $0.25. Fireworks serves the same model at $0.15. ([source](https://novita.ai/pricing))

### Context window

- 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))
- Novita AI: 1M tokens on the DeepSeek V4 line and GLM 5.3, published as "1M" rather than an exact figure — so it is recorded at 1,000,000 and may in practice be the 1,048,576 other providers state. Qwen3.8 is listed at 977K. ([source](https://novita.ai/pricing))

### H100 SXM, per GPU-hour

- 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))
- Novita AI: $3.39 per GPU-hour for an on-demand H100 SXM 80GB, or $1.70 spot. The page names the card the site's GPU comparison is fixed on, so no substitution is involved. ([source](https://novita.ai/gpus))

### Largest GPU offered

- Fireworks AI: 180GB per GPU on a B200, at $10.00 an hour on demand. ([source](https://fireworks.ai/pricing))
- Novita AI: 80GB per GPU on an H100 SXM, the largest card it lists; below it are an L40S 48GB, RTX 5090 32GB and RTX 4090 24GB. ([source](https://novita.ai/gpus))

## Which should you choose?

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.

Pick Novita AI if Breadth — 200+ models under one API with a cache-read rate published next to most of them, and an H100 SXM at $3.39 an hour if the same vendor should run the dedicated side too.

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.

Consider something else: Novita AI — It is rarely the cheapest for any given model: DeepInfra serves DeepSeek-V4-Flash-0731 at $0.06 against Novita's $0.44, and its H100 is $2.20 against Novita's $3.39 on demand.

## 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 Fireworks AI or Novita AI?**

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.
Pick Novita AI if Breadth — 200+ models under one API with a cache-read rate published next to most of them, and an H100 SXM at $3.39 an hour if the same vendor should run the dedicated side too.