D

DeepSeek V4 Flash

DeepSeek · Released Jul 2026

Recommended

DeepSeek's volume-tier open-weight model at 284B parameters (13B active MoE) with a 1M-token context window. Delivers SWE-bench Verified 79.0% at $0.14/$0.28 per 1M tokens — cheaper than any Western frontier-adjacent API by 7-107x. MIT-licensed for self-hosting on a single H100 or 2x A100s. The strongest value proposition in AI — near-frontier coding at commoditized pricing — use as the default workhorse for high-throughput agent pipelines; reserve Pro for the hardest reasoning tasks.

Is it right for you?

Good for

  • High-throughput pipelines — classification, extraction, summarization, and structured-output tasks at the lowest cost in frontier-adjacent AI
  • Cost-sensitive coding and agentic workflows — SWE-bench Verified 79.0% at $0.14/$0.28, beating models 10-100x its price
  • Self-hosted deployments on modest hardware — fits on a single H100 or 2x A100s at INT4 quantization

Not good for

  • Hardest reasoning and multi-step agentic tasks — V4 Pro ($0.435/$0.87) or Western flagships lead on sustained complex reasoning
  • Vision/multimodal workloads — text-only model

Pricing

Input

$0.14 / 1M tokens

Output

$0.28 / 1M tokens

Context

1M tokens

View full pricing

Benchmarks

BenchmarkScoreSource
SWE-bench Verified79.0% Source
SWE Pro (Resolved)52.6% Source
GPQA Diamond89.4% Source
τ²-Bench95.0% Source
Intelligence Index40.3 Source
Coding Index56.2 Source

No verdict changes yet

The clock starts day one — changes land here as our verdict evolves.

Verification log

  • Pricing— No changes

    Automated agent

How we evaluate