← Model indexMODEL PROFILE / deepseek-v4-1-flash

DeepSeek / current

DeepSeek V4.1 Flash

DeepSeek's MIT-licensed V4.1 Flash build tuned for software engineering benchmarks.

Context1M
ModalityText
AccessWeights · API
ReasoningHybrid

Where it fits.

Pairs open weights with a one-million-token context and reported strong results on coding and terminal benchmarks. Released October 17, 2026 as V4.1-Flash-1017 under an MIT license with a 1M-token context. Reported results include 77.0 on DeepSWE and 42.3 on Terminal-Bench, with Search and Fetch APIs metered at $0.75 per 1,000 calls billed from October 31, 2026.

02

Core capabilities

  • MIT-licensed open weights with 1M context
  • Strong coding and terminal benchmark results
  • Search and Fetch API tooling
  • OpenAI Chat Completions compatible interface
03

Best-fit workflows

  • Agentic coding
  • Terminal-driven workflows
  • Cost-sensitive open deployments
04 / OPERATIONS

From evaluation
to production.

Deployment notes

A strong open choice for coding agents with predictable licensing. Keep model IDs in configuration, cap output aggressively, validate tool arguments before execution, and meter API calls separately from tokens.

Watch before choosing

Search and Fetch APIs are billed per call from October 31, 2026, creating a second cost meter beyond tokens. Tool-call meters stack on top of token pricing; budget both. Very long outputs still need truncation controls and schema validation.

05

Family & versions

Specifications and positioning checked against provider documentation.

Open official source↗