Inference provider pricing comparison
Last updated: 2026-11-09
These prices were copied from each provider's official pricing pages, model pages and documentation on 2026-11-09. All prices are in USD per 1M tokens unless noted. They change often, so confirm on the provider's site before you commit. We did not run cost or performance tests ourselves.
gpt-oss-120b: the same model across providers
| Provider | Input | Cached input | Output | Vendor-listed speed |
|---|---|---|---|---|
| Baseten | $0.10 | Not listed | $0.50 | Not listed |
| Together AI | $0.15 | Not listed | $0.60 | Not listed |
| Fireworks AI | $0.15 | $0.015 | $0.60 | Fast mode: 100+ tokens/sec |
| Groq | $0.15 | Not listed | $0.60 | ~500 tokens/sec |
| Nebius Token Factory | $0.15 | $0.10 | $0.60 | Base config: 40 tokens/sec shown |
| SambaNova Cloud | $0.22 | Not listed | $0.59 | Not listed |
| Cerebras | $0.35 | Not listed | $0.75 | ~3,000 tokens/sec |
Cerebras announced gpt-oss-120b at $0.25 / $0.69 in August 2025. Its current model page lists $0.35 / $0.75. Groq also lists gpt-oss-20b at $0.075 / $0.30 (~1,000 tokens/sec). Nebius lists gpt-oss-20b from $0.05 input.
GLM-5.3
| Provider | Input | Cached input | Output |
|---|---|---|---|
| Together AI | $1.40 | $0.26 | $4.40 |
| Fireworks AI | $1.40 | $0.26 | $4.40 |
| Baseten | $1.40 | Not listed | $4.40 |
| Baseten (GLM-5.3 Fast) | $2.10 | Not listed | $6.60 |
| Nebius Token Factory | $1.40 | Not listed | $4.40 |
| DeepInfra (list price) | $0.90 | $0.20 | $4.00 |
| DeepInfra (temporary 38% discount shown) | $0.5625 | $0.125 | $2.50 |
Kimi K3
| Provider | Input | Cached input | Output |
|---|---|---|---|
| Together AI | $3.00 | $0.30 | $15.00 |
| Fireworks AI | $3.00 | $0.30 | $15.00 |
| Baseten | $3.00 | Not listed | $15.00 |
| DeepInfra | $2.85 | $0.285 | $14.25 |
Low-cost options on smaller models
- DeepInfra: Llama 3.1 8B Turbo $0.02 / $0.04, Llama 3.3 70B Turbo $0.10 / $0.32, DeepSeek-V4-Flash $0.09 / $0.18.
- Fireworks AI: NVIDIA Nemotron 3.5 Lightning 30B A3B $0.05 / $0.20.
- Nebius Token Factory and Baseten: GLM-5.3-Flash $0.15 / $0.50.
- SambaNova Cloud: Llama 3.3 70B $0.60 / $1.20; gemma-4-31B $0.38 / $1.15.
Discounts and service modes
| Provider | Mode | Published terms |
|---|---|---|
| Fireworks AI | Batch | 50% off input and output |
| Fireworks AI | Spot | 30% off Standard on select models; throughput can drop to zero |
| Fireworks AI | Priority | Higher rates (for example, GLM 5.3 at $1.75 / $5.50) |
| Fireworks AI | Prompt caching | Cached input at 50% off or more; no write fee |
| Together AI | Batch API and Provisioned Throughput | Batch pricing is shown per model; Provisioned Throughput was announced at $0.05 per throughput unit per minute with a 99% uptime SLA |
| Groq | Flex and Batch processing | Offered as service tiers; see Groq docs for terms |
| Nebius Token Factory | Cached tokens | Automatic; billed at each model's cached-input price on Serverless only |
| Baseten | Cached tokens | Automatic discounted rate for reused prompt prefixes |
Dedicated and per-second GPU rates
| Provider | H100 | B200 | Billing |
|---|---|---|---|
| Together AI (dedicated inference) | $5.49/hr | $8.99/hr | On-demand, per GPU-hour |
| Together AI (GPU clusters) | $3.99/hr | $8.19/hr | On-demand; reserved discounts available |
| Fireworks AI | $8.00/hr | $13.00/hr | Per GPU-second; region-restricted deployments cost 1.5x |
| Modal | $0.001097/sec (about $3.95/hr) | $0.001736/sec (about $6.25/hr) | Per second; region selection 1.15–1.75x; non-preemptible 3x |
| Replicate | $5.49/hr | Not listed | Per second |
| Nebius Token Factory | Shown per model in the console | Per GPU-hour, per-minute granularity | |
Gateway fees
- OpenRouter: 5.5% platform fee on Standard and 8% on Business, with discounts on Enterprise. Bring-your-own-key is free up to $25,000 of list-price inference per month, then 5%.
- Hugging Face Inference Providers: provider prices passed through with no markup. $2/month in credits on PRO and $2 per seat on Team or Enterprise.
Back to the ranked list of the best inference providers · Methodology
Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.