novita/deepseek-v3.2| Provider | Input, $ per 1M | Output, $ per 1M | In our data since |
|---|---|---|---|
| crof | $0.18 | $0.35 | 31 Jul 2026 |
| meganova | $0.26 | $0.38 | 31 Jul 2026 |
| DeepInfra | $0.26 | $0.38 | 31 Jul 2026 |
| Kilo Code | $0.26 | $0.38 | 31 Jul 2026 |
| LLM Gateway | $0.26 | $0.38 | 31 Jul 2026 |
| Cortecs | $0.266 | $0.444 | 31 Jul 2026 |
| novita-ai | $0.269 | $0.4 | 31 Jul 2026 |
| Novita AI | $0.269 | $0.4 | 31 Jul 2026 |
OpenRouter
deepseek/deepseek-v3.2
|
$0.269 | $0.4 | 31 Jul 2026 |
| alibaba-cn | $0.27 | $0.42 | 31 Jul 2026 |
| siliconflow-cn | $0.27 | $0.42 | 31 Jul 2026 |
| Abacus.AI | $0.27 | $0.4 | 31 Jul 2026 |
| Helicone | $0.27 | $0.41 | 31 Jul 2026 |
| Poe | $0.27 | $0.4 | 31 Jul 2026 |
| SiliconFlow | $0.27 | $0.42 | 31 Jul 2026 |
| vivgrid | $0.28 | $0.42 | 31 Jul 2026 |
| gmi | $0.28 | $0.4 | 31 Jul 2026 |
| DeepSeek | $0.28 | $0.4 | 31 Jul 2026 |
| Hugging Face | $0.28 | $0.4 | 31 Jul 2026 |
NanoGPT
deepseek/deepseek-v3.2
|
$0.28 | $0.42 | 31 Jul 2026 |
OpenRouter
openrouter/deepseek/deepseek-v3.2
|
$0.28 | $0.4 | 31 Jul 2026 |
| Vercel | $0.28 | $0.42 | 31 Jul 2026 |
| ZenMux | $0.28 | $0.43 | 31 Jul 2026 |
| 302.AI | $0.29 | $0.43 | 31 Jul 2026 |
| tensorx | $0.3 | $0.5 | 2 Aug 2026 |
| Nebius Token Factory | $0.3 | $0.45 | 31 Jul 2026 |
| Venice AI | $0.33 | $0.48 | 31 Jul 2026 |
| friendli | $0.5 | $1.5 | 31 Jul 2026 |
NanoGPT
TEE/deepseek-v3.2
|
$0.5 | $1 | 31 Jul 2026 |
| Fireworks AI | $0.56 | $1.68 | 1 Aug 2026 |
| Google Vertex AI | $0.56 | $1.68 | 31 Jul 2026 |
| Azure Cognitive Services | $0.58 | $1.68 | 31 Jul 2026 |
| Microsoft Foundry | $0.58 | $1.68 | 31 Jul 2026 |
| SambaNova | $3 | $4.5 | 31 Jul 2026 |
Only the base tier and only per-token prices are shown. Batch, discounted and cached rates, as well as prices per image or per second of video, do not go into this table: they cannot stand in the same column as a price per million tokens.
A single provider appears several times if it sells this model under different identifiers and at different prices — for example, at different weight precision. The identifier is shown next to the name. Identical offers that arrived from two sources under different spellings are merged into one row.
The date in the last column is the day this price first entered our collection. The price may well be older: before that day we simply were not recording it. It has not changed since — otherwise a new row with a new date would stand in its place.
| Task set | Result | Run conditions | Measured by |
|---|---|---|---|
| Arena Score, programming | 1,448.64 1,442–1,455 | — | — |
| Arena Score, mathematics | 1,436.18 1,425–1,447 | — | — |
| Arena Score in English | 1,435.55 1,431–1,440 | — | — |
| Arena Score, expert questions | 1,435.41 1,425–1,446 | — | — |
| Arena Score, hard prompts | 1,433.5 1,429–1,438 | — | — |
| Arena Score in French | 1,431.79 1,413–1,451 | — | — |
| Arena Score, long queries | 1,428.59 1,423–1,435 | — | — |
| Arena Score, multi-turn dialogue | 1,425.67 1,419–1,433 | — | — |
| Arena Score, overall | 1,424.39 1,421–1,428 | — | — |
| Arena Score in Russian | 1,423.66 1,415–1,432 | — | — |
| Arena Score in Spanish | 1,422.11 1,404–1,440 | — | — |
| Arena Score, instruction following | 1,412.51 1,407–1,418 | — | — |
| Arena Score, creative writing | 1,399.13 1,391–1,407 | — | — |
| Arena Score, web development | 1,323.38 1,315–1,331 | — | — |
| ARC-AGI — generalising to unseen patterns | 57 % | · with a tuned harness | — |
| Terminal-Bench — working in the command line | 39.6 % | · with a tuned harness | — |
| APEX-Agents | 7 % | — | — |
| ARC-AGI-2 | 4.03 % | — | — |