DeepSeek V4 Flash 0731
Historical record: DeepSeek retired this version on 2026-09-10
As of 2026-10-20, DeepSeek V4 Flash 0731 is historical: it entered public beta on 2026-07-31 and was replaced by V4.1-Flash on 2026-09-10; V4.1-Flash-1017 updated on 2026-10-17 with the same API and pricing, under the model name deepseek-flash on DeepSeek's own API. QCode's /models lists deepseek-v4-flash and deepseek-v4.1-flash; the exact version is governed by /models.
Updated 2026-10-22
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Highlights
Same API name at the time
0731 kept the existing Flash id, so no client rename was needed; since 2026-09-10 that id is retired and routed to V4.1-Flash.
Context / max out
Official pricing page. Thinking effort: low / high / max.
Cache miss / output
Cache hit $0.0028. Peak/off-peak from 2026-08-16 16:00 UTC.
Terminal Bench 2.1
Official: Harness Minimal + max. DeepSWE 54.4 (Preview was 7.3).
What 0731 actually is
On 2026-10-17 DeepSeek upgraded V4.1-Flash to DeepSeek-V4.1-Flash-1017: the API and pricing are unchanged and the model name remains deepseek-flash. DeepSeek says the new version improves benchmarks including HLE without tools at 42.6, Terminal-Bench 4.0 at 42.3, DeepSWE v1.1 at 77.0 and NL2Repo-Bench at 68.2, while shortening task runtime and output length. What DeepSeek officially said about the 2026-07-31 Flash post-training at the time was this: same Flash structure and scale, re-done post-training; HF lists about 304B (community shorthand often cites 284B/13B active MoE); MIT weights; the checkpoint ships DSpark speculative decoding. On 2026-09-10 DeepSeek released V4.1-Flash and retired the deepseek-v4-flash api name — those requests are served by V4.1-Flash and billed at the Flash price. So this is a historical comparison, not the current version.
2026-10-17: the V4.1-Flash-1017 update
On 2026-10-17 DeepSeek released DeepSeek-V4.1-Flash-1017: the API and pricing are unchanged and the model name remains deepseek-flash. DeepSeek says it improves benchmarks including HLE without tools at 42.6, Terminal-Bench 4.0 at 42.3, DeepSWE v1.1 at 77.0, ProgramBench at 29.1, NL2Repo-Bench at 68.2 and Automation-Bench v1.0.6 at 60.5, while shortening task runtime and output length. Reporting and official benchmarks at the time (2026-07-31): the preview Flash felt mediocre, 0731's AA approached the March 2026 frontier (about 50–52) and made it OpenCode's volume leader; LocalLLaMA called it "five-month-old frontier you can run at home". All of that is the 0731 narrative, not the current availability — the live Flash tier is V4.1-Flash-1017, updated on 2026-10-17 and succeeding V4.1-Flash from 2026-09-10.
Timeline
2026-04-24 V4-Pro / V4-Flash entered the API (preview).
2026-07-31 public beta of Flash 0731. Changelog: Flash only, naming the Responses API and Codex.
2026-09-10 DeepSeek released V4.1-Flash; deepseek-v4-flash was retired and routed to DeepSeek-V4.1-Flash, billed at the Flash price (pricing footnote (1)).
Confirmed facts vs claims to treat carefully
Confirmed
The benchmarks above are 0731's own claims at the time; 1M/384K; MIT; three effort levels; native Responses API; evaluation ran in DeepSeek Harness Minimal. DeepSeek's 2026-09-10 release note and changelog confirm deepseek-v4-flash and deepseek-v4-flash-vision-exp are retired, served by DeepSeek-V4.1-Flash and billed at the Flash price. On 2026-10-17 DeepSeek upgraded V4.1-Flash to DeepSeek-V4.1-Flash-1017: the API and pricing are unchanged and the model name remains deepseek-flash; official benchmarks are HLE without tools 42.6, Terminal-Bench 4.0 42.3, DeepSWE v1.1 77.0, ProgramBench 29.1, NL2Repo-Bench 68.2 and Automation-Bench v1.0.6 60.5.
Claims worth treating with caution
AA 52 against Sonnet 5's 55 is an independent evaluation, not an official "we beat Sonnet". The "196GB" in the blog is not a spec sheet. And "0731 is still today's Flash" is not true either: DeepSeek retired it on 2026-09-10 and updated to V4.1-Flash-1017 on 2026-10-17.
Quick compare
vs Claude Sonnet 5
AA 52 vs 55. Cost/task ~$0.06 vs $1.54. Flash is faster and open. Sonnet wins instruction stack and Claude Code defaults.
vs V4-Pro 0813
0731 beat Preview Pro. After 8-13, 0813 is the flagship (TB2.1 87.9, DeepSWE 62.7). Flash stays the default cheap lane.
How to call it
Official api.deepseek.com: the model name deepseek-flash points to the latest V4.1-Flash-1017; the 0731-era deepseek-v4-flash is still routed for compatibility. Also an Anthropic-compatible path and Responses/Codex docs. Local: HF deepseek-ai/DeepSeek-V4-Flash-0731 with vLLM/SGLang DSpark. OpenCode Go/Free also list Flash.
On QCode
0731 is the historical version; see the official docs for the 1017 update. QCode's /models lists deepseek-v4-flash and deepseek-v4.1-flash; the exact version is governed by /models. For new integrations use deepseek-v4.1-flash, or deepseek-v4-pro for the Pro tier. Billing is per token; per-model rates are on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.
Flash 0731 FAQ
Is 0731 a new model?
Same Flash structure with new post-training, and the API name was unchanged for the 2026-07-31 release. But since 2026-09-10 DeepSeek retired deepseek-v4-flash and routes it to V4.1-Flash; on 2026-10-17 it updated to V4.1-Flash-1017, so this only describes 0731 at the time.
Can I run it locally?
Yes, MIT weights. Full size needs serious unified memory or multi-GPU; 16GB laptops want a quant, not the full checkpoint.
Is it better than Sonnet 5?
On AA intelligence, slightly behind (52 vs 55). On price and speed, far ahead. Pick by stack, not a trophy.
Does 8-16 pricing kill the deal?
0731 priced peak output at $1.32 and off-peak at $0.66. That is still well under Sonnet 5's $10/$15 band. The 2026-10-17 V4.1-Flash-1017 keeps pricing unchanged: cache hit $0.003 off-peak / $0.006 peak, cache miss $0.15 / $0.3 and output $0.6 / $1.2 per 1M tokens.
What did the V4.1-Flash-1017 update change?
On 2026-10-17 DeepSeek upgraded V4.1-Flash to DeepSeek-V4.1-Flash-1017: the API and pricing are unchanged and the model name remains deepseek-flash. DeepSeek reports higher benchmarks (HLE without tools 42.6, Terminal-Bench 4.0 42.3, DeepSWE v1.1 77.0, NL2Repo-Bench 68.2) and shorter task runtime and output length.
Flash or Pro?
Default to Flash (concurrency 2500). Escalate hard agent jobs to Pro 0813 (concurrency 500). V4.1-Flash-1017 from 2026-10-17 is still called as deepseek-flash on DeepSeek's own API with the same API and pricing. On QCode use deepseek-v4.1-flash; which version it maps to is governed by /models.
Sources
DeepSeek's official changelog, the news260910 and news261017 release notes, and Models & Pricing (fetched 2026-09-18, re-checked 2026-10-20); the 0731-era official blog and third-party evaluations. Retirement, routing and the 1017 update follow the changelog and release notes.
Put Flash in a multi-model mix
0731 is the historical record. For new integrations use deepseek-v4.1-flash from QCode's /models, or deepseek-v4-pro for the Pro tier; the exact version is governed by /models. Billing is per token; per-model rates are on /models.
Related
Flash vs Sonnet 5
Full AA scoreboard.
V4-Pro 0813
The flagship that followed.
DeepSeek Harness
The official Minimal shell behind the benches.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
This page keeps the 2026-07-31 official benchmarks and third-party numbers from that time without claiming the version is still available. V4.1-Flash replaced it on 2026-09-10; the live Flash tier is V4.1-Flash-1017, updated on 2026-10-17; defer to DeepSeek's docs and QCode's /models. Checked 2026-10-20.