Live pricing on OneInfer
Pulled from the OneInfer catalog at request time — rates can change. Capture the timestamp with any production decision.
| Provider | Input $/1M tokens | Output $/1M tokens | Cache read $/1M tokens |
|---|---|---|---|
| novita | $2.00 | $6.00 | $0.250 |
| together_ai | $2.00 | $6.00 | $0.250 |
Published API pricing per 1M tokens
Source: each provider's own listing, fetched 8 September 2026. One row per provider serving Qwen3.8-2.4T-A95B on OneInfer.
| Provider | Input $/1M | Output $/1M | Cache read $/1M | Batch input / output | Verified at |
|---|---|---|---|---|---|
| Alibaba Qwen (direct) | $2.00 | $6.00 | n/a | $1.00 / $3.00 | 8 Sep 2026 |
| DeepInfra | $2.00 | $6.00 | $0.20 | n/a | 8 Sep 2026 |
| Novita | $2.00 | $6.00 | $0.25 | n/a | 8 Sep 2026 |
| Vercel AI Gateway | $2.00 | $6.00 | $0.25 | n/a | 8 Sep 2026 |
| Hugging Face (routes Novita) | $2.00 | $6.00 | n/a | n/a | 8 Sep 2026 |
| Together AI | $2.50 | $6.25 | $0.50 | n/a | 8 Sep 2026 |
Together AI charges 25% more
Together AI is the only tracked provider with a price premium: 25% more on input, 4% more on output, and 2.5× the cache-read rate of DeepInfra. The premium is consistent with Together AI's other frontier MoE listings, but is the single biggest per-token cost swing across the six tracked providers.
Alibaba batch cuts cost in half
Alibaba Cloud's batch endpoints run at $1.00 input / $3.00 output per 1M tokens — exactly half the standard rate. For offline reprocessing jobs the batch endpoint is the cheapest published route.
Ready to test the workflow?
How blended cost is calculated
Blended cost assumes one input token for every three output tokens (a 1:3 input:output ratio): ($2.00 + 3 × $6.00) ÷ 4 = $5.00 per 1M tokens at the $2.00 / $6.00 list rate. At 1:1 the blend is $4.00, and at an input-heavy 3:1 input:output ratio it is $3.00. Treat blended as a workload sensitivity test, not a neutral sort key.
Cheapest-first ordering
Alibaba Qwen batch ($1.00 / $3.00) → DeepInfra cache-heavy ($0.20 cache read) → five at-parity providers ($2.00 / $6.00) → Together AI ($2.50 / $6.25, $0.50 cache). Fireworks supports serverless inference but its published rate was not captured at this page's verification.