DeepSeek model product
DeepSeek-V3
A powerful Mixture-of-Experts (MoE) language model with 671B total parameters (37B activated per token).
Updated Aug 12, 2026. Default version: DeepSeek-V3 0324
LLMBoard score26.0DeepSeek-V3 0324
Coverage80%5 benchmark families
Context window163.8KTokens
Official input priceN/AOfficial price unavailable
DeepSeek-V3 Specifications
Technical details for the model's default version.
- Version
- DeepSeek-V3 0324
- Released
- Mar 25, 2025
- Knowledge cutoff
- Unknown
- Parameters
- 671B
- Context window
- 163.8K
- Max output
- 163.8K
- Inputs
- text
- Outputs
- text
- Open weights
- No
- License
- MIT + Model License (Commercial use allowed)
DeepSeek-V3 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
DeepSeek-V3 Benchmark Results
Benchmark scores for DeepSeek-V3 0324.
| MATH-500 | 94.0% | 22 | 32 | 32.3% | C | |
| AIME 2024 | 59.4% | 44 | 53 | 17.3% | C | |
| MMLU-Pro | 81.2% | 44 | 129 | 66.4% | C | |
| LiveCodeBench | 49.2% | 47 | 73 | 36.1% | C | |
| GPQA | 68.4% | 136 | 234 | 42.1% | C |
DeepSeek-V3 Arena Results
Preference and agent-evaluation results for the default version.
| text style control | overall | 142 | 1395.7 | 45,626 | N/A | |
| text | overall | 152 | 1375.1 | 45,626 | N/A |
DeepSeek-V3 Runtime Performance
Provider-specific output speed and catalog latency for DeepSeek-V3 0324. Runtime does not affect the capability score.
No runtime data
Browse runtime rankingsNo provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
DeepSeek-V3 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.20 input, $0.77 output per 1M via NanoGPT
- Tracked offerings
- 7
| NanoGPT | deepseek-v3-0324 | global | $0.20 | $0.77 | 128K | |
| submodel | deepseek-ai/DeepSeek-V3-0324 | global | $0.20 | $0.80 | 75K | |
| Meganova | deepseek-ai/DeepSeek-V3-0324 | global | $0.25 | $0.88 | 163.8K | |
| NovitaAI | deepseek/deepseek-v3-0324 | global | $0.27 | $1.12 | 163.8K | |
| OpenRouter | deepseek/deepseek-chat-v3-0324 | global | $0.27 | $1.12 | 163.8K | |
| Jiekou.AI | deepseek/deepseek-v3-0324 | global | $0.28 | $1.14 | 163.8K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
DeepSeek-V3 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| DeepSeek-V3 0324 | 26.0 | 671B | 163.8K | 163.8K | No | MIT + Model License (Commercial use allowed) | |
| DeepSeek-V3 | N/A | 671B | 131.1K | 8.2K | Yes | MIT + Model License (Commercial use allowed) |
What is DeepSeek-V3?
Key information about DeepSeek-V3 and its available data.
A powerful Mixture-of-Experts (MoE) language model with 671B total parameters (37B activated per token). Features Multi-head Latent Attention (MLA), auxiliary-loss-free load balancing, and multi-token prediction training. 8T tokens with strong performance in reasoning, math, and code tasks.
Data as of 2026-08-11.
DeepSeek-V3 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to DeepSeek-V3
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
DE
DeepSeek-R1-Distill-Qwen
DeepSeek
DE
DeepSeek-V2.5
DeepSeek
DE
DeepSeek-R1-Distill-Llama
DeepSeek
DE
DeepSeek-V3.1
DeepSeek
DE
DeepSeek-R1
DeepSeek
DE
DeepSeek-VL2
DeepSeek
FAQ
Common questions about DeepSeek-V3.
When was DeepSeek-V3 released?
DeepSeek-V3's default version was released on Mar 25, 2025.
How much does DeepSeek-V3 cost?
No official standard PAYG price is currently available for DeepSeek-V3. The lowest tracked third-party offer starts at $0.20 input and $0.77 output via NanoGPT.
Who created DeepSeek-V3?
DeepSeek-V3 was created by DeepSeek.
What is the context window for DeepSeek-V3?
The default version has a 163.8K token context window.
Is DeepSeek-V3 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer DeepSeek-V3?
7 provider offerings are linked to the default version.
What models should I compare DeepSeek-V3 with?
Nearby ranked alternatives include Qwen3 VL 8B Thinking, Gemini 2.0 Flash, Phi 4 Reasoning Plus.