DeepSeek V3
by DeepSeekChat
DeepSeek chat model for instruction following, coding, and analysis
- Context
- 164K tokens
- Max output
- 16,000 tokens
- Parameters
- 685B total, 37B active
- Released
- 2024-12-26
- Weights
- Open
deepseek/deepseek-v3 · also seen as deepseek-ai/DeepSeek-V3, public/deepseek-v3, Pro/deepseek-ai/DeepSeek-V3, deepinfra/deepseek-ai/DeepSeek-V3, +2 more
Specifications
Full facts as recorded, incl. what is unknown.
- Released
- 2024-12-26
- Weights
- Opennot gated
- Input
- Text
- Output
- Text
- Context window
- 164,000 max across endpoints
- Max output
- 16,000
- Providers
- 18 + 7 gateways, not counted as providers
- Epoch Capabilities Index
- 132.3 ± 4.2 against price
- Direct price range
- $0.19–$1.25 in · $0.38–$1.50 out per 1M tokens · 11 of 18 providers have a comparable price
More details
- Last updated
- 2025-11-25
- Knowledge cutoff
- 2024-07
- Weights link
- Hugging Face · deepseek-ai/DeepSeek-V3
- Parameters
- 685B total, 37B active
- Developer country
- China
- Training compute
- 3.3×10²⁴ FLOP
Capabilities
Across all endpoints: supported if any endpoint supports it.
- Tool calling
- Parallel tools
- Structured output
- Reasoning
- Image input
- PDF input
- Audio input
- Audio outputunknown
- Video input
- Prompt caching
- Web searchunknown
- Computer useunknown
- System messages
- Assistant prefill
- Temperature
- Streaming
- Attachments
Endpoints
One row per provider and variant; gateways are listed separately.
Loading endpoints…
Price history
Atom feedNo change since 19 Oct 2026. The cheapest direct price has been $0.19 in / $0.38 out per 1M tokens at inferenceapis on every data refresh since then.
Tracked from 19 Oct 2026, the earliest refresh this model's headline price can be rebuilt from. Headline-eligible direct prices only, so a price the sanity check flags never draws a spike. Series as JSON
Cost calculator
Loading endpoints…
Similarly priced models
Blended rate per 1M tokens, each at its own cheapest provider (10M input : 2M output mix).
| Model | Blended/1M | In/1M | Out/1M | Context | vs DeepSeek V3 |
|---|---|---|---|---|---|
| Qwen3-Vl 30b A3b ThinkingAlibaba (Qwen) | $0.408 | $0.29 | $1.00 | 262K | +0% |
| MiniMax-M2.1MiniMax | $0.408 | $0.27 | $1.10 | 205K | +0% |
| Llama Guard 3 8BMeta | $0.408 | $0.484 | $0.03 | 131K | −0% cheaper |
| Hermes 3 70B InstructNous Research | $0.408 | $0.408 | $0.408 | 131K | −0% cheaper |
| DeepSeek V4 ProDeepSeek | $0.406 | $0.348 | $0.696 | 1M | −1% cheaper |
| GLM-4.6Z.ai (Zhipu AI) | $0.411 | $0.274 | $1.10 | 205K | +1% |
| GLM-4.7Z.ai (Zhipu AI) | $0.411 | $0.274 | $1.10 | 205K | +1% |
| GLM-4.5VZ.ai (Zhipu AI) | $0.411 | $0.274 | $1.10 | 64K | +1% |
More from DeepSeek
Newest DeepSeek models in the catalogue. All DeepSeek models
| Model | Released | Context | Listed price |
|---|---|---|---|
| DeepSeek V4 Flash | 2026-10-17 | 1.05M | from $0.07 |
| DeepSeek V4.1 Flash 1017 | 2026-10-17 | 1M | from $0.08 |
| DeepSeek V4 Flash Vision Exp | 2026-10-17 | 1.05M | from $0.14 |
| DeepSeek V4.1 Flash | 2026-09-10 | 1.05M | from $0.09 |
| DeepSeek V4 Pro 0813 | 2026-08-12 | 1M | from $0.88 |
| DeepSeek V4 Flash Latest | 2026-07-31 | 1.05M | from $0.04 |
| DeepSeek Pro Latest | 2026-07-06 | 1M | unpriced |
| DeepSeek Latest | 2026-05-03 | 1.05M | $0.145 via NanoGPT |