// DeepSeek
DeepSeek V3
A DeepSeek model. 9 prices across 6 channels side by side, each with its source and check date – plus specs, licence, speed and benchmarks.
- Official in/out per 1M
- —
- Lowest (GMICloud)
- $0.209 / $0.31
- Context
- 164K
- AA index
- 8.5
Prices checked 2026-10-22Official sources linked on every rowNo estimatesCheapest LLM APIs →
01Specs at a glance
Official values first; gaps filled from OpenRouter’s API where the vendor page is silent.
| Vendor | DeepSeek |
|---|---|
| Context window | 164K (163,840 tokens) |
| Max output | 147K |
| Input | Text |
| Output | Text |
| Tool calling / JSON | Yes / Yes |
| Weights & licence | Open weights · MIT · Hugging Face |
| Parameters | 685B (from weight files) |
| Released | Dec 1, 2025 |
| Model IDs | deepseek/deepseek-v3.2 · deepseek-v3.2 · deepseek/deepseek-chat · deepseek-ai/DeepSeek-V3.2 |
02Prices on 6 channels
Per 1M tokens, in the currency each source publishes (1 USD = 6.7 CNY for the blended column). Sorted: official first, then cheapest.
| Channel | Source | Input | Output | Cache hit | Blended $ | Context | Notes |
|---|---|---|---|---|---|---|---|
| GMICloud · fp8 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.209 | $0.31 | $0.022 | 0.234 | 164K | GMICloud price as listed on OpenRouter (pass-through; quantization=fp8; discount field=0.28); not checked against GMICloud's own page |
| DeepInfra Inference platform (own price page) | Platform 10-10 | $0.26 | $0.38 | $0.13 | 0.29 | 164K | DeepInfra own pricing page |
| AtlasCloud · fp8 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.26 | $0.38 | $0.13 | 0.29 | 164K | AtlasCloud price as listed on OpenRouter (pass-through; quantization=fp8; discount field=0); not checked against AtlasCloud's own page |
| SiliconFlow · fp8 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.259 | $0.8 | $0.14 | 0.394 | 164K | SiliconFlow price as listed on OpenRouter (pass-through; quantization=fp8; discount field=0); not checked against SiliconFlow's own page |
| DeepInfra · fp4 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.32 | $0.89 | — | 0.463 | 164K | DeepInfra price as listed on OpenRouter (pass-through; quantization=fp4; discount field=0); not checked against DeepInfra's own page |
| SiliconFlow Inference platform (own price page) | Platform 10-10 | ¥4 | ¥6 | ¥0.4 | 0.672 | — | SiliconFlow (硅基流动) own pricing page |
| SiliconFlow Inference platform (own price page) | Platform 10-10 | ¥4 | ¥6 | ¥0.4 | 0.672 | — | SiliconFlow (硅基流动) own pricing page |
| Friendli · unknown Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.5 | $1.5 | $0.25 | 0.75 | 164K | Friendli price as listed on OpenRouter (pass-through; quantization=unknown; discount field=0); not checked against Friendli's own page |
| SambaNova · unknown Via OpenRouter / provider listed there | OpenRouter 10-22 | $3 | $4.5 | — | 3.38 | 33K | SambaNova price as listed on OpenRouter (pass-through; quantization=unknown; discount field=0); not checked against SambaNova's own page |
“Official” = vendor pricing page; “Platform” = the platform’s own pricing page; “OpenRouter” = price listed on OpenRouter, not checked on the platform’s site. Click a source tag to open the original page.
03Where you can use it
Derived only from the price sources above. Country restrictions are set by each vendor – check their terms before you sign up.
- Via OpenRouter / provider listed there
GMICloud · AtlasCloud · SiliconFlow · DeepInfra · Friendli · SambaNova - Inference platform (own price page)
DeepInfra · SiliconFlow - Self-host the open weights
GPU / VRAM sizing and build plans→ - Free option
DeepSeek-V3.2→
04Performance
All results →Speed and AA index from Artificial Analysis public leaderboards (not our own tests); benchmark scores from our benchmark library, best result per benchmark.
- AA indexDeepSeek V3 (Dec '24)
- 8.5
- Output speed (median)
- 41 tok/s
- Time to first token
- 1.18s
- AA index per $ (blended)
- 36.27
| Benchmark | Best score | Source | Date |
|---|---|---|---|
| LMArena Text | 1425 | official | 2026-10-08 |
| LMArena WebDev | 1362 | official | 2026-10-08 |
| GPQA Diamond | 83.4% | independent | 2025-12-16 |
| MMLU-Pro | 85.0% | vendor | 2026-03-11 |
| SWE-bench Verified | 70.0% | official | 2026-02-17 |
| AIME (OTIS Mock AIME 2024–2025) | 87.8% | independent | 2025-12-16 |
| τ²-bench | 96.2% | vendor | 2025-12-01 |
| Terminal-Bench 2.0 | 39.5% | official | 2025-12-01 |
05Price changes we recorded
All price changes →From our daily re-checks of the sources above.
No price change recorded since we started tracking (2026-10-10).
06More from DeepSeek
- DeepSeek V4.1 Flash 1017
$0.098 / 0.294 · 1.05M→ - DeepSeek V4.1 Flash
$0.14 / 0.42 · 1M→ - DeepSeek V4 Flash Vision Exp
$0.216 / 0.647 · 1.05M→ - DeepSeek-V4-Flash
$0.06 / 0.18 · 1.05M→ - DeepSeek-V4-Pro
$0.957 / 1.91 · 1M→ - DeepSeek V3.2 Exp
$0.27 / 0.41 · 164K→ - DeepSeek V3.1 Terminus
$0.27 / 1 · 164K→ - DeepSeek-V3.1
$0.25 / 0.95 · 164K→
FAQ
How much does the DeepSeek V3 API cost?
There is no official pay-as-you-go price in our data. The lowest listed price is $0.209 / $0.31 on GMICloud.
What is the context window of DeepSeek V3?
164K tokens of context, up to 147K output tokens. Platforms may cap it lower – see the table.
Is DeepSeek V3 open source?
Yes, the weights are published on Hugging Face (deepseek-ai/DeepSeek-V3.2). Licence: MIT.