DeepSeek, released Mar 24, 2025

DeepSeek V3 0324price, context, benchmarks and release details

Provisional: not enough results to rank yet Open weights 74% confidence 74 percent, Medium confidence, 4 of 9 expected sources in
55.7
SI Score
#82 of 142 ranked models
Input, per 1M tokens
not yet reported
Output, per 1M tokens
not yet reported
Context window
164Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Max output
164Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released
Mar 24, 2025models.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗

How the score breaks down

Coding (weight 40 percent) 48.5
Math (weight 15 percent) 63.0
Preference (weight 15 percent) 70.5
Reasoning (weight 30 percent) 67.6

Weights: reasoning 30%, math 15%, coding 40%, preference 15%. Results use fixed 0–100 scales before averaging, and thin evidence is pulled toward 50. Method si-v3-retained-evidence-2, computed Oct 9, 2026, 05:13 UTC.

Around it on the leaderboard

  1. 80 Fugu 55.9
  2. 81 MiMo-V2.6-Flash 55.8
  3. 82 DeepSeek V3 0324 55.7
  4. 83 GPT-5 55.7
  5. 84 MiniMax-M2 55.7

Full leaderboard

Benchmark results

6 benchmarks, 6 results

Each row shows the best published result. Where a model was tested at several settings, such as reasoning effort, open the row to see each one. Hover or tap a value for its source.

GPQA Diamondreasoning 67.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Apr 1, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 67.6
MATH Level 5math 75.5%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Apr 1, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 75.5
OTIS Mock AIME 2024–2025math 37.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Apr 1, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 37.8
Aider Polyglotcoding 55.1%Aider polyglotPublished source fact [variant] Aider polyglot; 225 cases; 2 attemptsPublished Mar 24, 2025 Retrieved Oct 9, 2026 · Apache-2.0
Open source ↗ 55.1
SWE-bench Verifiedcoding 42%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] SWE-ExpPublished Aug 6, 2025 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 42.0
LMArena Textpreference 1374.6 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 8, 2026 Retrieved Oct 9, 2026 · CC-BY-4.0
Open source ↗ 70.5

Normalization uses a fixed 0–100 scale for each unit, independent of other models. Compare evaluation conditions before reading a small gap as decisive. “Lab-reported” marks the provider's own published figure.

Details and sources

Open weights
Yesmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
License
not yet reported
Input modalities
textmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
First seen by SuperIndex
Oct 8, 2026
Coverage
59% of expected source weight

Reported (4)

  • Aider polyglotOct 8, 2026
  • Epoch AI BenchmarkingOct 8, 2026
  • LMArena / ArenaOct 8, 2026
  • SWE-bench VerifiedOct 8, 2026

Awaiting (5)

  • ARC Prize11% of weight
  • Humanity’s Last Exam11% of weight
  • Official model cards via models.dev4% of weight
  • LiveBench11% of weight
  • Terminal-Bench5% of weight

Confidence rises as pending sources publish. Some sources never cover some models, so confidence reaches 100% at 80% of expected weight.

What changed

All releases · RSS feed

Alerts on this device

What to be alerted about
RSS feed