DeepSeek V3
DeepSeek-V3 is a 671B-parameter Mixture-of-Experts language model from DeepSeek with 37B activated parameters per token, trained on 14.8T tokens.
- Input / output price
- $0.32 / $0.89 per 1M
- Context
- 164K tokens
- Released
- Dec 26, 2024
Paste into Claude Code, Cursor, or Codex
Help me set up the Respan gateway: fetch and follow https:/
Providers
3 providers, cheapest first, priced per 1M tokens after gateway discounts.
| Provider | Input | Output |
|---|---|---|
Azure DeepSeek azure_deepseek/deepseek-v3 | $0.14/M | $0.28/M |
DeepInfra deepinfra/deepseek-v3 | $0.32/M | $0.89/M |
Novita novita/deepseek-v3 | $0.89/M | $0.89/M |
Pricing
Averaged across 3 providers, in USD per 1M tokens after gateway discounts.
- Average input
- $0.45 per 1M
- Average output
- $0.69 per 1M
- Average cache read
- $0.04 per 1M
Cost by volume
Total cost from 0 to 1T tokens on the cheapest and most expensive provider.
Throughput and latency
30-day averages from requests through the Respan gateway. Under each figure, every model on the gateway by its average, this one in blue.
Output speed
Higher is better6.3tokens/sacross providers
Response latency
Lower is better0.8sacross providers
Benchmarks
The Respan Index plus 11 published scores from 6 sources, each linked to where it was published.
13.2/ 100
Respan Index#167 of 176
- 167DeepSeek V313.2
- 166gpt-oss-20b13.6
- 168Llama 4 Scout11.2
Published scores
| Benchmark | Score | Source | Published |
|---|---|---|---|
| SWE-bench Verifiedgithub.com | 42% | github.com | Dec 26, 2024 |
| WeirdMLHåvard Tveit Ihle | 36.1% | Håvard Tveit Ihle | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| GPQA Diamondgithub.com | 59.1% | github.com | Dec 26, 2024 |
| GPQA DiamondEpoch AI | 56.5% | Epoch AI | Oct 31, 2026 |
| SimpleBenchSimpleBench | 18.9% | SimpleBench | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| MATH Level 5Epoch AI | 64.9% | Epoch AI | Oct 31, 2026 |
| AIME 2025MathArena | 25% | MathArena | Sep 17, 2026 |
| OTIS Mock AIME 2024-2025Epoch AI | 15.8% | Epoch AI | Oct 31, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| MMLUgithub.com | 88.5% | github.com | Dec 26, 2024 |
| MMLU-Progithub.com | 75.9% | github.com | Dec 26, 2024 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| LMArena EloLMArena | 1358 Elo · #150 of 191 | LMArena | Nov 19, 2026 |
How DeepSeek compares: Claude vs DeepSeek, DeepSeek vs Qwen, DeepSeek vs Gemini
Compare DeepSeek V3: DeepSeek V3 vs DeepSeek V3.1, DeepSeek V3 vs DeepSeek-V4.1-Flash-1017, DeepSeek V3 vs DeepSeek V4.1 Flash, DeepSeek V3 vs DeepSeek V4 Pro, DeepSeek V3 vs DeepSeek V4 Flash, DeepSeek V3 vs DeepSeek-V4-Flash (Preview)
Apps
Use DeepSeek V3 in the tools you already run. Each guide points the app at the Respan gateway.
Setup guides for more apps are in the integration docs.
DeepSeek V3 FAQ
Short answers, built from the data on this page.
DeepSeek V3 is a model from Deepseek, released on Dec 26, 2024. DeepSeek-V3 is a 671B-parameter Mixture-of-Experts language model from DeepSeek with 37B activated parameters per token, trained on 14.8T tokens.
On the Respan gateway, DeepSeek V3 costs from $0.14 per 1M input tokens and $0.28 per 1M output tokens on Azure DeepSeek, after gateway discounts. Averaged across its 3 providers, it costs $0.45 per 1M input tokens and $0.69 per 1M output tokens.
DeepSeek V3 is available through 3 providers: Azure DeepSeek, DeepInfra, and Novita.
DeepSeek V3 accepts up to 164K tokens of context on its largest provider route.
Point any OpenAI-compatible SDK at https://gateway.respan.ai/v1 with your Respan API key and set the model to "deepseek-v3". To set it up with a coding agent, copy the agent setup prompt at the top of this page.
DeepSeek V3 scores 13.2 out of 100 on the Respan Index v1.1, ranking #167 of 176 models. By category: Coding 3.1, Agents 5.6 (estimated), Reasoning 17.3, Knowledge & language 7.9, and Human preference 39.4. The Respan Index combines independent and lab-reported benchmark results into one score, using benchmark data collected by Respan.
DeepSeek V3 has 11 published benchmark scores, including SWE-bench Verified 42%, GPQA Diamond 59.1%, and MATH Level 5 64.9%. Each score links to where it was published; the data is collected by Respan.
Certified where it counts.
Built to meet the security and privacy standards that enterprise and healthcare teams require.
ISO 27001
The internationally recognized standard for information security management.
SOC 2
Secure, compliant management of your data across all of our systems.
GDPR
Operated under GDPR, the world's strictest standard for data privacy.
HIPAA
HIPAA compliant, with a BAA available for healthcare teams.