Skip to content
OSMEXO
Sign inGet an API key
☰
Sign inGet an API key

DeepSeek V4.1 Flash 1017

deepseek/deepseek-v4.1-flash-1017
  • DeepSeek
  • Family · DeepSeek V4 Flash
  • Released 17 Oct 2026
  • Open weights
  • Text · Image → Text
  • Reasoning · adjustable
  • Tools · JSON · Vision
Live on OSMEXO
Try nowRequest a discount
API setup & price →
Spec & list price
Context1M
Max output384K
Input / 1M$0.15
Output / 1M$0.604× input
Scores · standing
Intelligence45.1#20/489
LiveBench72.9#9/32
Usage
Tokens / day6.0T▲ 38% 7d
Usage rank#1
Requests / day65.2M
Tool calls / req0.48
Per request · 30 days
Avg input60.6K4.1× median
Avg output8561.2× median
Cache hit94%1.3× median
Reasoning share68%1.2× median
Tool-call failures1.4%0.7× median

Consumption

3.3T6.7T10.0T26 Oct29 Oct1 Nov
Peak 8.2T 30 OctLow 682.8B 17 OctRank today #1

News & trends

62 mentions this week ▲ 100%this model · videos, posts, articles · 30 days
Trending · 14 days
News · latest from DeepSeek V4 Flash

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis
Long-context reasoning#25 of 393
82.7%
SciCode#29 of 160
55.9%
Intelligence index#20 of 489
45.1
Humanity's Last Exam#34 of 465
43.3%
Terminal-Bench 4.0#20 of 152
33.8%
Speed · median across 42 providers, latest day
Output speed#44/334126 tok/s1.9× faster than median
Fastest provider282 tok/sbest of 42 providers
First token#1400.89 s1.2× quicker than median
First answer token#5014 sthinking included

#2 cheapest of 13 with a similar score

Value
Intelligence Index
3035404550$0.1$0.2$0.5$1$2$5DeepSeek V4.1 Flash 1017
Blended price per 1M tokens (log)
Family progress
Intelligence Index
20304050MayJunJulAugSepOctDeepSeek V4.1 Flash 1017
  • Family releases45.1
LiveBench suites
Browser67.2
Coding96.0
Data Analysis79.5
Design83.4
Long Agentic35.7
Reasoning89.1
Research79.5

▲ Strongest: Coding, Reasoning. ▼ Weakest: Long Agentic (36).

Technical details

Reasoning
on · high
Tokenizer
DeepSeek
License
MIT
Parameters
763B
Moderation
none
Downloads · likes
177.0K · 4.6K
Reasoning effort
  • max
  • high · default
  • low
  • none
Request parameters 22
  • tools
  • tool_choice
  • parallel_tool_calls
  • response_format
  • structured_outputs
  • reasoning
  • reasoning_effort
  • include_reasoning
  • temperature
  • top_p
+12 more
  • top_k
  • min_p
  • max_tokens
  • max_completion_tokens
  • stop
  • seed
  • frequency_penalty
  • logit_bias
  • logprobs
  • presence_penalty
  • repetition_penalty
  • top_logprobs

Interesting facts

  1. Less output room than similar models (393.2K against a 734.0K median).
  2. Released 16 days ago, the newest of 4 DeepSeek V4 Flash models.
  3. 2nd of 4 DeepSeek V4 Flash models by tokens last month.

When to use

Best for

  • Coding assistants and code generation at low cost
  • Reasoning-heavy tasks over very long context
  • Screenshot and image understanding with text answers
  • Browser-style agents with tool calling
  • Self-hosted deployments (open weights)

Think twice for

  • Long autonomous agent runs (weakest suite)
  • Image generation or editing
  • Audio or video input

Research papers

Same family

DeepSeek V4 Flash

Alternatives

other vendors · similar capability

Using DeepSeek V4.1 Flash 1017 at volume?

Sign in to ask for a volume discount. You will continue in the OSMEXO dashboard.

Request a discount