Home › VastLLM Catalog › DeepSeek V4 Flash
DeepSeek V4 Flash
by DeepSeek · Language Models
Context 1M tokensReasoning / thinkingFunction calling
by DeepSeek · Language Models
DeepSeek V4 Flash is the efficiency-optimized member of DeepSeek's V4 family, released open-source under the MIT license on April 24, 2026. Per its model card, it is a Mixture-of-Experts model with 284B total and 13B active parameters, sharing the V4 hybrid CSA plus HCA attention architecture and a reported 1M-token context window. It supports Non-think, Think High and Think Max reasoning modes and posts strong agentic and coding results, with reported scores of 79.0% on SWE-Bench Verified and 56.9% on Terminal-Bench 2.0. Ideal for fast, high-throughput, cost-sensitive workloads needing solid reasoning and coding.
$0.2849 in · $0.8547 out / 1M
| Usage | Rate (USD) | Unit |
|---|---|---|
| Input tokens | $0.28 | per 1M tokens |
| Cached input (cache hit) | $0.01 | per 1M tokens |
| Output tokens | $0.85 | per 1M tokens |
TuneRouter live platform pricing · currency: USD · pre-deduction settled by actual usage
OpenAI-compatible · base URL https://api.tunerouter.com/v1/
curl https://api.tunerouter.com/v1/chat/completions \
-H "Authorization: Bearer $TUNEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "deepseek-v4-flash", "messages": [{"role": "user", "content": "Hello!"}]}'Model ID: deepseek-v4-flash · works with OpenAI SDKs — just point base_url at api.tunerouter.com