Platform comparison
Novita AI vs Together AI: GPT-OSS 120B compared
Choose Novita AI for GPT-OSS 120B at $0.05/$0.25 per 1M input/output tokens and 228 tokens/s in the Artificial Analysis (AA) snapshot; Together AI has the shorter first-chunk median, 0.47 s versus 0.91 s.
| Dimension | Winner | Margin | Evidence |
|---|---|---|---|
| Output speed | Novita AI | ~1.9× | 228 vs 118 tokens/s3 |
| Median first-chunk latency | Together AI | 0.44 s | 0.47 vs 0.91 s3 |
| Price (input / output) | Novita AI | ~67% / ~58% | $0.05 / $0.25 vs $0.15 / $0.6012 |
Performance
Performance
AA provider-table snapshot, 2026-10-15: GPT-OSS 120B at high reasoning effort and 10,000 input tokens.
Swipe to see all columns
| Metric | Novita AI | Together AI |
|---|---|---|
| Median output speed (tokens/s) | 228 | 118 |
| First-chunk median (s) | 0.91 | 0.47 |
| Endpoint Accuracy Index v1.0 (point estimate) | 93% | 81% |
Pricing
Pricing
GPT-OSS 120B token prices in USD per 1M tokens.
Swipe to see all columns
| Metric | Novita AI | Together AI |
|---|---|---|
| Input / 1M tokens | $0.05 | $0.15 |
| Output / 1M tokens | $0.25 | $0.60 |
Capabilities and serving conditions
How each platform runs in production
| Dimension | Novita AI | Together AI |
|---|---|---|
| API & SDK | GPT-OSS 120B uses openai/gpt-oss-120b at https://api.novita.ai/openai.1 | GPT-OSS 120B is available as openai/gpt-oss-120b through the Together API and its Python and TypeScript SDKs.2 |
| Model catalog | Novita advertises 200+ models across text, image, video and audio; the Batch comparison below applies specifically to GPT-OSS 120B.8 | Together advertises 200+ models; GPT-OSS 120B is a text-input, text-output model with a listed 128K context.29 |
| Batch processing | Batch accepts up to 50,000 requests per file within a 24-hour processing window. The listed 50% discount is introductory.4 | Batch supports up to 50,000 requests and 100 MB per input file. Its 24-hour target is best-effort; GPT-OSS 120B is not in the current discounted-model list.5 |
| Serving options | Dedicated Endpoints scale to zero when idle and bill per GPU-second. The token prices above apply to the serverless API.7 | Serverless, dedicated model inference and provisioned throughput are separate options. Serverless can return 429 or 503 under high demand; committed capacity requires a separate agreement.6 |
Which to choose
Choose by token budget and capacity requirements
Choose Novita AI if…
- A 100M-input/20M-output workload has estimated base token charges of $10 on Novita AI versus $27 on Together AI.
- Long generated responses benefit from the observed 228 tokens/s versus Together AI’s 118 tokens/s; this does not remove Novita AI’s longer first-chunk delay.
- You want to combine GPT-OSS 120B inference with Novita image, video, audio and Agent Sandbox APIs in one provider account.
Choose Together AI if…
- A shorter initial wait matters more than sustained output rate: AA records 0.47 s versus 0.91 s for the first chunk.
- Your selected model or model family needs contracted provisioned throughput. Together’s published serverless service is best-effort rather than a throughput guarantee.
- A separate asynchronous queue fits offline processing, but GPT-OSS 120B currently keeps standard token rates in Together’s Batch model list.
Methodology
Methodology and sources
AA's 2026-10-15 GPT-OSS 120B provider table supplies the Novita and Together AI performance cells: high reasoning effort, 10,000 input tokens and 72-hour medians. The first received chunk may contain reasoning; output speed measures the subsequent token rate. Endpoint Accuracy Index v1.0 compares each endpoint with AA's reference, rather than reporting a task pass rate or a statistically established advantage.
The Together AI comparison prices a fixed 100M-input and 20M-billed-output budget at uncached synchronous rates, including any charged reasoning tokens. The estimates exclude tax, retries, caching, Batch discounts and contracted capacity. Serverless list prices do not represent a provisioned-throughput quote. Model counts retain the two platforms' advertised catalog scopes. Novita AI prepared this public-source comparison without running a private workload benchmark.
- Novita AI · GPT-OSS 120B novita.ai · Checked 2026-10-15
- Together AI · GPT-OSS 120B www.together.ai · Checked 2026-10-15
- Artificial Analysis · GPT-OSS 120B Providers artificialanalysis.ai · Checked 2026-10-15
- Novita AI · Batch API docs.novita.ai · Checked 2026-10-15
- Together AI · Batch Overview docs.together.ai · Checked 2026-10-15
- Together AI · Serverless Rate Limits docs.together.ai · Checked 2026-10-15
- Novita AI · Dedicated Endpoints docs.novita.ai · Checked 2026-10-15
- Novita AI · Model APIs novita.ai · Checked 2026-10-15
- Together AI · Model Library www.together.ai · Checked 2026-10-15
Build with Novita Model APIs
Create an API key and connect GPT-OSS 120B to your application.