SI Index Data as of Oct 8, 2026

DeepSeek V4.1 Flash: Price, Context, Benchmarks, and Release Details

DeepSeek open weights · first seen Oct 8, 2026 · score computed Oct 8, 2026, 22:13 UTC
66.2
SI Score (method si-v2-absolute-shrinkage-1)
#27 of 115 ranked
69% confidence 69 percent, Medium confidence — 4 of 7 expected sources in
Coverage 55% of expected source weight · 100% confidence at 80% coverage

Pillar breakdown

Coding (weight 40 percent) 73.3
Math (weight 15 percent) 88.0
Preference (weight 15 percent) 78.8
Reasoning (weight 30 percent) 82.7

Pillar weights: reasoning 30% · math 15% · coding 40% · preference 15%. Benchmark results use fixed 0–100 scales before averaging; incomplete evidence is shrunk toward 50.

Facts

Input price / 1M
$0.15DeepSeek pricingOfficial off-peak uncached rate; peak is 2x; time schedule at source Retrieved Oct 8, 2026 · factual citation
Open source ↗
Output price / 1M
$0.60DeepSeek pricingOfficial off-peak uncached rate; peak is 2x; time schedule at source Retrieved Oct 8, 2026 · factual citation
Open source ↗
Context window
1Mmodels.devPublished source fact Retrieved Oct 8, 2026 · MIT
Open source ↗
Max output
384Kmodels.devPublished source fact Retrieved Oct 8, 2026 · MIT
Open source ↗
Released
Sep 10, 2026DeepSeek V4.1 Flash announcementOfficial dated introduction and availability announcement Retrieved Oct 8, 2026 · factual citation
Open source ↗
Open weights
YesHugging Face HubPublic Hub repo with weight files; gating/repo upload date is not release date Retrieved Oct 8, 2026 · factual metadata; model-specific licenses
Open source ↗
License
mitHugging Face HubPublished source fact Retrieved Oct 8, 2026 · factual metadata; model-specific licenses
Open source ↗
Input modalities
text, imagemodels.devPublished source fact Retrieved Oct 8, 2026 · MIT
Open source ↗
Popularity rank
#1 by usageOpenRouter rankingsPublished source fact; tokens processed; page default period; excludes catalog and endpoint statistics Retrieved Oct 8, 2026 · CC-BY-4.0
Open source ↗

Benchmark results

43 results
Benchmark Raw result Normalized (0–100) Pillar
arc agi v1 public eval 95.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-highPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		95.9 reasoning
arc agi v1 public eval 96.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort using the standard harness. [variant] deepseek-v4-1-flash-lowPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		96.3 reasoning
arc agi v1 public eval 98%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-maxPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		98.0 reasoning
arc agi v1 semi private 88.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-highPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		88.5 reasoning
arc agi v1 semi private 90.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort using the standard harness. [variant] deepseek-v4-1-flash-lowPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		90.5 reasoning
arc agi v1 semi private 94.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-maxPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		94.5 reasoning
arc agi v2 public eval 71.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-highPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		71.9 reasoning
arc agi v2 public eval 66.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort using the standard harness. [variant] deepseek-v4-1-flash-lowPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		66.8 reasoning
arc agi v2 public eval 81.7%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-maxPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		81.7 reasoning
arc agi v2 semi private 67.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-highPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		67.5 reasoning
arc agi v2 semi private 60.6%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort using the standard harness. [variant] deepseek-v4-1-flash-lowPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		60.6 reasoning
arc agi v2 semi private 72.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. [variant] deepseek-v4-1-flash-maxPublished Oct 6, 2026 Retrieved Oct 8, 2026 · factual citation
Open source ↗ 		72.9 reasoning
gpqa diamond 90.9%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		90.9 reasoning
hle full set 36.8%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; full set Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		36.8 reasoning
hle text only subset 39.1%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; text-only subset Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		39.1 reasoning
hle tools 63.9%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; with tools Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		63.9 reasoning
livebench coding code completion 2026_06_25 82.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		82.6 coding
livebench coding code generation 2026_06_25 77.5%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		77.5 coding
livebench coding javascript 2026_06_25 81.8%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		81.8 coding
livebench coding python 2026_06_25 90%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		90.0 coding
livebench coding typescript 2026_06_25 60%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		60.0 coding
livebench math amps hard 2026_06_25 98%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		98.0 math
livebench math integrals with game 2026_06_25 90%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		90.0 math
livebench math math comp 2026_06_25 96.1%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		96.1 math
livebench math olympiad 2026_06_25 89.1%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		89.1 math
livebench math simplify 2026_06_25 67.1%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		67.1 math
livebench reasoning connections 2026_06_25 100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		100.0 reasoning
livebench reasoning consecutive events 2026_06_25 88.7%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		88.7 reasoning
livebench reasoning logic with navigation 2026_06_25 68%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		68.0 reasoning
livebench reasoning spatial 2026_06_25 98%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		98.0 reasoning
livebench reasoning theory of mind 2026_06_25 80.8%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		80.8 reasoning
livebench reasoning zebra puzzle 2026_06_25 100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 8, 2026 · factual citation; Apache-2.0 code
Open source ↗ 		100.0 reasoning
lmarena text 1462.5 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 2, 2026 Retrieved Oct 8, 2026 · CC-BY-4.0
Open source ↗ 		78.8 preference
terminal bench v2 1 88%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; Claude Code; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		88.0 coding
terminal bench v2 1 84.1%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; Codex; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		84.1 coding
terminal bench v2 1 90.6%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; DeepSeek Harness minimal mode; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		90.6 coding
terminal bench v2 1 85.8%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; DeepSeek Harness PTC mode; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		85.8 coding
terminal bench v2 1 85.8%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; DeepSeek Harness standard mode; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		85.8 coding
terminal bench v2 1 90.3%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; mini-swe-agent; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		90.3 coding
terminal bench v2 1 85%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; OpenCode; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		85.0 coding
terminal bench v2 1 86.1%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; Pi; 2.1 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		86.1 coding
terminal bench v3 0 30%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; DeepSeek Harness minimal mode; 3.0 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		30.0 coding
terminal bench v4 0 31.2%Official model cards via models.devLab-reported; metric pass@1; transcribed by MIT models.dev catalog; not independently evaluated [variant] reasoning_effort=100; DeepSeek Harness minimal mode; 4.0 Retrieved Oct 8, 2026 · factual citation; MIT transcription
Open source ↗ 		31.2 coding

Normalization uses fixed absolute 0–100 scales for each unit, independently of other models that have a result on each benchmark. Coverage and evidence breadth still affect the composite; compare the evaluation conditions before reading a small score gap as decisive. “lab-reported” marks the provider's own published figure.

Sources: in and pending

in Reported (4)

  • ARC Prize · arrived Oct 8, 2026
  • Official model cards via models.dev · arrived Oct 8, 2026
  • LiveBench · arrived Oct 8, 2026
  • LMArena / Arena · arrived Oct 8, 2026

pending Awaiting (3)

  • Epoch AI Benchmarking · carries 25% of expected weight
  • Humanity’s Last Exam · carries 13% of expected weight
  • Terminal-Bench · carries 6% of expected weight

The confidence % rises as pending sources publish. Some sources never cover some models — that is why 100% confidence arrives at 80% of expected weight, not at full coverage.