Leaderboard · Model · Updated
DeepSeek V3
671B MoE open model with strong frontier-tier performance at very low API cost.
At a glance
| Provider | DeepSeek |
|---|---|
| Released | 2024-12 |
| Tier | open-weights |
| License | Open-weights (DeepSeek License) |
| Modalities | text |
| Context window | 128k tokens |
| Max output | 8.192k tokens |
| API price · input | $0.27 / 1M tokens |
| API price · output | $1.10 / 1M tokens |
| Hugging Face | deepseek-ai/DeepSeek-V3 |
Benchmark performance
How DeepSeek V3 stacks up against the current leader (GPT-5) and the median model in the leaderboard:
| Benchmark | DeepSeek V3 | GPT-5 | Median |
|---|---|---|---|
| Chatbot Arena Elo | 1320 | 1410 | 1320 |
| MMLU-Pro | 75.9 | 86.8 | 78.0 |
| GPQA Diamond | 59.1 | 87.3 | 65.0 |
| MATH | 90.2 | 96.7 | 78.3 |
| HumanEval | 91.0 | 95.1 | 92.0 |
| SWE-Bench Verified | 42.0 | 74.9 | 49.0 |
Numbers compiled from provider technical reports and Chatbot Arena. See methodology for the composite-score formula.
OpenRouter routes your requests across DeepSeek V3, GPT-5, Claude, Gemini, and 100+ other models behind a single API key — pay-as-you-go, no monthly minimum. Try OpenRouter → (affiliate · supports this site)
What does DeepSeek V3 cost in practice?
API pricing is $0.27 per 1M input tokens and $1.10 per 1M output tokens. Assuming a 50/50 input/output split, here is what that looks like at three workload sizes:
| Volume | Per day | Per month | Per year |
|---|---|---|---|
| 1M tokens/day | $0.69 | $20.55 | $250.03 |
| 10M tokens/day | $6.85 | $205.50 | $2,500 |
| 100M tokens/day | $68.50 | $2,055 | $25,003 |
Strengths & weaknesses
Where it shines
No clear top-tier strengths — this is a mid-pack model.
Where it lags
- SWE-Bench: 42.0 (rank #15 of 18, below average)
- Arena: 1320 (rank #17 of 27, mid-pack)
Best alternatives
The closest models to DeepSeek V3 by tier and benchmark score:
| Model | Score | $ in / out | Context | Action |
|---|---|---|---|---|
| Qwen2.5-Coder 32B Alibaba |
68.8 | $0.18 / $0.18 | 131.072k | Try → · vs DeepSeek |
| Llama 3.1 405B Instruct Meta |
65.7 | $2.70 / $2.70 | 128k | Try → · vs DeepSeek |
| Qwen2.5 72B Instruct Alibaba |
65.6 | $0.35 / $0.40 | 131.072k | Try → · vs DeepSeek |
| Phi-4 Microsoft |
71.2 | $0.07 / $0.14 | 16.384k | Try → · vs DeepSeek |
| Llama 3.3 70B Instruct Meta |
64.7 | $0.23 / $0.40 | 128k | Try → · vs DeepSeek |
Frequently asked questions
Is DeepSeek V3 a good model?
DeepSeek V3 scores 68.0 on the llmrank.top composite (rank #18 of 30). It trails the frontier — for top performance, look at GPT-5 (86.0). Whether it's the right fit depends on your workload — see the use-case discussion on this page.
How much does DeepSeek V3 cost?
DeepSeek V3 is priced at $0.27 per 1M input tokens and $1.10 per 1M output tokens (USD). For a 10M-token-per-day workload split 50/50, that works out to roughly $6.85/day or $2,500/year.
What is DeepSeek V3's context window?
128k tokens. For very long inputs, consider a 1M+ context model like Gemini 2.5 Pro or GPT-4.1.
Is DeepSeek V3 open source?
Yes — released under the DeepSeek License license, weights are downloadable from the provider or Hugging Face.
What is DeepSeek V3's SWE-Bench score?
DeepSeek V3 scores 42.0% on SWE-Bench Verified — the benchmark that measures real-world GitHub issue resolution. Its Chatbot Arena Elo is 1320.
What are the best alternatives to DeepSeek V3?
Closest alternatives by tier and score: Qwen2.5-Coder 32B, Llama 3.1 405B Instruct, Qwen2.5 72B Instruct. See the alternatives section on this page for side-by-side numbers.
Related: Qwen2.5-Coder 32B · Llama 3.1 405B Instruct · Qwen2.5 72B Instruct · Full leaderboard
Spotted out-of-date numbers? Open an issue — corrections usually ship within 24h.
Try DeepSeek V3 now
Direct link to the official playground — or use OpenRouter to A/B test it against any other model on a single API key.