# DeepSeek-V4-Flash (deepseek-reasoner)

> Thinking mode of DeepSeek-V4-Flash: 1M context, chain-of-thought reasoning and tool calling at low cost.

_The Frontier Model Matrix · /models/deepseek-reasoner · [JSON](/api/models/deepseek-reasoner) · [all The Frontier Model Matrix](/models)_

- **name:** DeepSeek-V4-Flash (deepseek-reasoner)
- **vendor:** DeepSeek
- **model_id:** deepseek-reasoner
- **context_window:** 1M
- **max_output:** 384K
- **input_per_mtok:** $0.14
- **output_per_mtok:** $0.28
- **strengths:** Thinking mode of DeepSeek-V4-Flash: 1M context, chain-of-thought reasoning and tool calling at low cost.
- **provider:** DeepSeek
- **family:** DeepSeek V4
- **release_date:** 2025-12-01 (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **last_updated:** 2026-06-15
- **open_weights:** true (source: https://api-docs.deepseek.com/quick_start/pricing)
- **license:** open (MIT-class; see model card) (source: https://api-docs.deepseek.com/quick_start/pricing)
- **params_total:** — (verify-against-primary-at-build)
- **params_active:** — (verify-against-primary-at-build)
- **tool_call:** true (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **reasoning:** always (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **structured_output:** — (verify-against-primary-at-build)
- **attachment:** — (verify-against-primary-at-build)
- **temperature:** true (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **knowledge_cutoff:** 2025-09 (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **context_advertised:** 1M (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **context_effective:** — (verify-against-primary-at-build)
- **price_input:** $0.14 (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **price_output:** $0.28 (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **price_cache_read:** — (verify-against-primary-at-build)
- **price_cache_write:** — (verify-against-primary-at-build)
- **cost_per_full_window:** — (verify-against-primary-at-build)
- **cost_per_agent_task:** — (verify-against-primary-at-build)
- **modalities:** text-in,text-out (source: https://models.dev/models/deepseek/deepseek-reasoner/)
- **gpqa_diamond:** — (verify-against-primary-at-build)
- **swe_bench_verified:** — (verify-against-primary-at-build)
- **terminal_bench:** — (verify-against-primary-at-build)
- **tau2_bench:** — (verify-against-primary-at-build)
- **bfcl_tool_use:** — (verify-against-primary-at-build)
- **aa_index:** — (verify-against-primary-at-build)
- **lmarena_elo:** — (verify-against-primary-at-build)
- **tokens_per_sec:** — (verify-against-primary-at-build)
- **ttft:** — (verify-against-primary-at-build)
- **hallucination_rate:** — (verify-against-primary-at-build)
- **agent_readiness_score:** — (verify-against-primary-at-build)
- **score_confidence:** partial
- **source_url:** https://models.dev/models/deepseek/deepseek-reasoner/
- **source_type:** models.dev
- **last_verified:** 2026-06-15
- **_provenance_note:** Context (1M), max output (384K) and price ($0.14/$0.28) cross-confirmed by DeepSeek's own pricing docs AND models.dev. deepseek-reasoner = thinking mode of DeepSeek-V4-Flash. structured_output not marked on the canonical entry -> placeholder.