DeepSeek · lightweight line · open weights
The newest DeepSeek Flash model is DeepSeek-V4.1-Flash-1017, released on October 17, 2026. It shipped 7 days ago, and the line's usual gap is about 29 days.
How long it has been since the last release against the line's usual gap and its recent ones — the rhythm, not a date.
Latest: DeepSeek-V4.1-Flash-1017 · Oct 17, 2026
Every Monday, only when something shipped: the AI models that came out, what each one changed, and which lines are due next.
We email you a link to confirm first · unsubscribe anytime · read past issues · about the newsletter · RSS for the DeepSeek Flash line
Newest first — what each version changed versus the one before, in DeepSeek's own numbers.
An update to DeepSeek-V4.1-Flash that scores higher on every agentic, coding and reasoning test DeepSeek published while giving shorter answers, with the same MIT weights, 1M-token context and API price.
Compared with DeepSeek-V4.1-Flash · the model it replaces
Completes more multi-step tasks at the command line
Terminal-Bench 4.0 — running real commands to complete a task
42.3
▲ up from 31.2 · +11.1 points
Scores higher than DeepSeek-V4.1-Flash on a combined set of hard tests
Artificial Analysis Intelligence Index v4.3.2 — one score averaging 10 hard tests
45
▲ up from 39 · +6 points · both at max effort
Costs developers the same as the model it replaces
Price per 1M tokens — what developers pay, input / output, at peak hours
$0.30 / $1.20
peak hours · same as DeepSeek-V4.1-Flash · $0.15 / $0.60 off-peak
A new encoder-decoder design that activates 8B parameters on input and 16B on output, beating V4-Flash-0731 on every agentic test DeepSeek published, with the MIT licence, a 1M-token context and a lower API price than the model it replaces.
Compared with DeepSeek-V4-Flash-0731 · the model it replaces
Finishes 9 in 10 multi-step tasks at the command line
Terminal-Bench 2.1 — running real commands to complete a task
90.6%
▲ up from 82.7% · +8 points
Scores 5 points higher on the all-round intelligence index
Artificial Analysis Intelligence Index v4.3.2 — one score averaging 10 hard tests
39
▲ up from 34 · +5 points
Costs developers less than the model it replaces
Price per 1M tokens — what developers pay, input / output, at peak hours
$0.30 / $1.20
▲ peak hours · down from $0.44 / $1.32 · input 32% cheaper, output 9% cheaper
The official release of V4-Flash keeps the preview's 284B-parameter design but, after further training, scores 82.7% on Terminal Bench 2.1, up from 61.8%, in DeepSeek's tests.
Compared with DeepSeek-V4-Flash-Preview · the preview it replaces
Finishes 8 in 10 multi-step tasks at the command line
Terminal-Bench 2.1 — running real commands to complete a task
82.7%
▲ up from 61.8% · +21 points
Scores 10 points higher on the all-round intelligence index
Artificial Analysis Intelligence Index v4.3.2 — one score averaging 10 hard tests
34
▲ up from 24 · +10 points
Launched at the same price as the preview
Price per 1M tokens — what developers pay, input / output
$0.14 / $0.28
same as the preview · $0.44 / $1.32 at peak hours from Aug 16
The smaller of the two V4 preview models, with 284B total and 13B active parameters, a 1M-token context window and MIT-licensed open weights, priced at $0.14 per million input tokens and $0.28 per million output tokens.
Solves about 8 in 10 real-world coding tasks
SWE-bench Verified — fixing real bugs in real software projects
79.0%
✦ first model in this line · DeepSeek's own test at max reasoning effort
Scores 24 on the all-round intelligence index
Artificial Analysis Intelligence Index v4.3.2 — one score averaging 10 hard tests
24
✦ first model in this line — nothing earlier to compare
Costs developers 14 cents per million input tokens
Price per 1M tokens — what developers pay, input / output
$0.14 / $0.28
✦ launch price — nothing earlier to compare