Sign Up

Gemini 3.8 Flash

Google DeepMind
Announcement
Gemini 3.8 Flash launch post cover

Cost/1M tokens

input
$0.825
cached
$0.0825
output
$4.125
cache write
$0.0417
Context
1Mtokens
Released
2mo ago

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Arena leaderboards and other sourced evaluations appear below; missing scores are not zero.

How it ranks

Against its predecessor, similarly priced models and the models ranked around it.

DeepSWE 1.1

Best pass@1 per model · Higher is better

#2 of 28

  1. 74.1%
    GPT-6 Astra
  2. 73.8%
    Gemini 3.8 Flash
  3. 73.6%
    Claude Opus 5
  4. 72.7%
    GPT-5.6 Sol
  5. 69.9%
    Claude Fable 5
  6. 69.6%
    GPT-5.6 Terra
  7. 69.0%
    GLM 5.3
  8. 68.5%
    Kimi K3
  9. 67.5%
    Grok 4.6
  10. 67.2%
    GPT-5.6 Luna

Arena Text

Best Arena rating per model · Higher is better

Arena #9

  1. 1503
    Claude Fable 5.1Nearby rank
  2. 1501
    Claude Opus 4.7Nearby rank
  3. 1497
    Gemini 3.8 Flash
  4. 1496
    Muse Spark 1.3Nearby rank
  5. 1493
    Muse Spark 1.2Nearby rank
  6. 1489
    Gemini 3.7 FlashPredecessor
  7. 1475
    GLM 5.2Similar price
  8. 1447
    GPT-5.4 MiniSimilar price
  9. 1443
    InklingSimilar price
  10. 1442
    Grok 4.3Similar price

Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs) · Ahead of 4 of 4 similarly priced models

Arena WebDev

Best Arena rating per model · Higher is better

Arena #42

  1. 1608
    GLM 5.2Similar price
  2. 1588
    Gemini 3.7 FlashPredecessor
  3. 1585
    MiniMax M3.1 FlashNearby rank
  4. 1583
    Gemini 3.8 Flash
  5. 1582
    DeepSeek V4 Flash 0423Nearby rank
  6. 1581
    DeepSeek V4 Pro 0813Nearby rank
  7. 1414
    InklingSimilar price
  8. 1398
    GPT-5.4 MiniSimilar price
  9. 1357
    Grok 4.3Similar price

Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs) · Ahead of 3 of 4 similarly priced models

Arena Vision

Best Arena rating per model · Higher is better

Arena #11

  1. 1294
    Gemini 3.7 FlashPredecessor
  2. 1294
    Muse Spark 1.2Nearby rank
  3. 1293
    Gemini 3.8 Flash
  4. 1293
    GPT-6.1 SolNearby rank
  5. 1291
    Muse Spark 1.3Nearby rank
  6. 1250
    GPT-5.4 MiniSimilar price
  7. 1245
    Step 5 PreviewSimilar price
  8. 1242
    Grok 4.3Similar price
  9. 1232
    GLM 5V TurboSimilar price

Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs) · Ahead of 4 of 4 similarly priced models

Arena Search

Best Arena rating per model · Higher is better

Not ranked on Arena Search yet

Arena Document

Best Arena rating per model · Higher is better

Arena #19

  1. 1480
    Muse Spark 1.1Similar price
  2. 1474
    GPT-6 AstraNearby rank
  3. 1471
    GPT-5.4Nearby rank
  4. 1469
    Gemini 3.8 Flash
  5. 1469
    Claude Sonnet 5Nearby rank
  6. 1468
    Gemini 3.7 FlashPredecessor
  7. 1450
    Kimi K2.6Similar price
  8. 1419
    Claude Haiku 4.5Similar price
  9. 1406
    GLM 5V TurboSimilar price

Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs) · Ahead of 3 of 4 similarly priced models

Agent Arena

Best IPS score per model · Higher is better

Arena #20

  1. +0.120
    GPT-6.1 TerraSimilar price
  2. +0.025
    Inkling 2 SmallNearby rank
  3. +0.024
    MiMo-V2.6-ProNearby rank
  4. +0.023
    Gemini 3.8 Flash
  5. +0.022
    Muse Spark 1.3Nearby rank
  6. +0.020
    DeepSeek V4.1 Flash 1017Nearby rank
  7. +0.002
    GLM 5.2Similar price
  8. -0.022
    Gemini 3.7 FlashPredecessor
  9. -0.037
    Step 5 PreviewSimilar price
  10. -0.173
    InklingSimilar price

+0.045 vs Gemini 3.7 Flash (predecessor) · Ahead of 3 of 4 similarly priced models

ARC-AGI-2

Best score per model · Higher is better

#10 of 85

  1. 89.6%
    GPT-6 SolNearby rank
  2. 89.2%
    Claude Fable 5Nearby rank
  3. 89.2%
    Gemini 3.8 Flash
  4. 88.3%
    Claude Opus 5Nearby rank
  5. 85.0%
    GPT-5.5Nearby rank
  6. 84.6%
    Gemini 3.7 FlashPredecessor
  7. 65.1%
    Grok 4.20Similar price
  8. 36.5%
    InklingSimilar price
  9. 22.8%
    GLM 5.2Similar price
  10. 18.9%
    GPT-5.4 MiniSimilar price

+4.6 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 4 of 4 similarly priced models

FrontierMath-Tier-4-v2-Private

Best score per model · Higher is better

#47 of 72

  1. 36.6%
    Gemini 3.7 FlashPredecessor
  2. 29.3%
    GLM 5.2Similar price
  3. 24.4%
    Grok 4.5Nearby rank
  4. 22.0%
    Gemini 3.6 FlashNearby rank
  5. 22.0%
    Gemini 3.8 Flash
  6. 22.0%
    GPT-5Nearby rank
  7. 19.5%
    GPT-5 ProNearby rank
  8. 14.6%
    Grok 4.3Similar price
  9. 9.8%
    GPT-5.4 MiniSimilar price
  10. 4.9%
    InklingSimilar price

-14.6 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 3 of 4 similarly priced models

FrontierMath-Tiers-1-3-v2-Private

Best score per model · Higher is better

#31 of 87

  1. 71.6%
    Gemini 3.7 FlashPredecessor
  2. 70.2%
    Claude Opus 4.7Nearby rank
  3. 68.8%
    GLM 5.3Nearby rank
  4. 68.4%
    Gemini 3.8 Flash
  5. 67.4%
    GPT-5.2Nearby rank
  6. 67.4%
    DeepSeek V4.1 FlashNearby rank
  7. 59.2%
    GLM 5.2Similar price
  8. 51.2%
    GPT-5.4 MiniSimilar price
  9. 42.8%
    Grok 4.3Similar price
  10. 33.3%
    InklingSimilar price

-3.2 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 4 of 4 similarly priced models

GPQA diamond

Best score per model · Higher is better

#5 of 161

  1. 95.8%
    GPT-6 AstraNearby rank
  2. 95.6%
    Claude Sonnet 5.5Nearby rank
  3. 95.4%
    Gemini 3.8 Flash
  4. 95.4%
    GPT-6.1 SolNearby rank
  5. 94.8%
    Gemini 3.7 FlashPredecessor
  6. 91.9%
    GLM 5.2Similar price
  7. 88.8%
    Grok 4.3Similar price
  8. 88.3%
    InklingSimilar price
  9. 86.9%
    GPT-5.4 MiniSimilar price

+0.6 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 4 of 4 similarly priced models

Humanity's Last Exam

Best score per model · Higher is better

#5 of 34

  1. 46.4%
    Gemini 3.1 Pro PreviewNearby rank
  2. 46.1%
    Claude Sonnet 5.5Nearby rank
  3. 44.5%
    Gemini 3.8 Flash
  4. 44.3%
    GPT-5.4 ProNearby rank
  5. 37.5%
    Gemini 3 Pro PreviewNearby rank
  6. 24.4%
    Kimi K2.5Similar price
  7. 11.0%
    Gemini 2.5 Flash Preview 05-20Earlier version
  8. 8.3%
    GLM 4.5Similar price

+33.6 pts vs Gemini 2.5 Flash Preview 05-20 (earlier version) · Ahead of 2 of 2 similarly priced models

OTIS Mock AIME 2024-2025

Best score per model · Higher is better

#19 of 158

  1. 98.9%
    Claude Haiku 5.5Nearby rank
  2. 98.9%
    Claude Opus 5Nearby rank
  3. 98.9%
    Gemini 3.8 Flash
  4. 98.9%
    GPT-6 LunaNearby rank
  5. 98.6%
    DeepSeek V4 Pro 0813Nearby rank
  6. 97.2%
    Gemini 3.7 FlashPredecessor
  7. 93.3%
    Grok 4.3Similar price
  8. 88.9%
    GPT-5.4 MiniSimilar price
  9. 88.9%
    InklingSimilar price
  10. 86.4%
    GLM 5.2Similar price

+1.7 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 4 of 4 similarly priced models

SimpleQA Verified

Best score per model · Higher is better

#10 of 87

  1. 70.8%
    Claude Fable 5.1Nearby rank
  2. 70.7%
    Claude Fable 5Nearby rank
  3. 69.7%
    Gemini 3.8 Flash
  4. 69.7%
    GPT-5.6 SolNearby rank
  5. 69.2%
    Gemini 3.7 FlashPredecessor
  6. 40.3%
    InklingSimilar price
  7. 34.2%
    GLM 5.2Similar price
  8. 33.2%
    Grok 4.3Similar price
  9. 29.4%
    GPT-5.4 MiniSimilar price

+0.5 pts vs Gemini 3.7 Flash (predecessor) · Ahead of 4 of 4 similarly priced models

DeepSWE 1.1

Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22

Official leaderboard
73.8%
pass@1 · highest published configuration
mini-swe-agent · high reasoning
pass@4
85.8%
Average cost / attempt
$2.36
Benchmark runs
4

Score vs cost by reasoning effort

Average cost per attempt (USD) against pass rate

All published configurations

Harness / reasoningpass@1pass@4Cost / attempt
mini-swe-agent · high73.8%85.8%$2.36
mini-swe-agent · medium71.0%83.2%$1.97

pass@1 is the success rate across scored attempts; pass@4 is the share of tasks solved in at least one of four attempts. Costs are reported by DeepSWE for each configuration. These results are not directly comparable to evaluations using different harnesses. Official source JSON.

Reasoning effort

ARC-AGI-2

Score by reasoning effort · Higher is better

All published scores

  • Arena TextArena #91497

    Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs)

    • high149795% CI 1493 to 1502 · 61,143 votes
    Source:Arena Text leaderboard, 2026-11-09
  • Arena WebDevArena #421583

    Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs)

    • high158395% CI 1576 to 1589 · 13,847 votes
    Source:Arena WebDev leaderboard, 2026-11-10
  • Arena VisionArena #111293

    Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs)

    • high129395% CI 1285 to 1301 · 7,624 votes
    Source:Arena Vision leaderboard, 2026-11-09
  • Arena DocumentArena #191469

    Level with Gemini 3.7 Flash (predecessor, overlapping 95% CIs)

    • high146995% CI 1461 to 1477 · 6,950 votes
    Source:Arena Document leaderboard, 2026-11-09
  • Agent ArenaArena #20+0.023

    +0.045 vs Gemini 3.7 Flash (predecessor)

    • high+0.02395% CI +0.014 to +0.032 · 8,602,306 observations
    Source:Agent Arena leaderboard, 2026-11-11
  • ARC-AGI-2#10 of 8589.2%

    +4.6 pts vs Gemini 3.7 Flash (predecessor)

    • low77.5%
    • medium82.9%
    • high89.2%
    Source:External evaluation via Epoch AIOriginal-source terms apply
  • FrontierMath-Tier-4-v2-Private#47 of 7222.0%

    -14.6 pts vs Gemini 3.7 Flash (predecessor)

    • high22.0%
    Source:Epoch AI evaluation
  • FrontierMath-Tiers-1-3-v2-Private#31 of 8768.4%

    -3.2 pts vs Gemini 3.7 Flash (predecessor)

    • high68.4%
    Source:Epoch AI evaluation
  • GPQA diamond#5 of 16195.4%

    +0.6 pts vs Gemini 3.7 Flash (predecessor)

    • high95.4%
    Source:Epoch AI evaluation
  • Humanity's Last Exam#5 of 3444.5%

    +33.6 pts vs Gemini 2.5 Flash Preview 05-20 (earlier version)

    • unknown44.5%
    Source:External evaluation via Epoch AIOriginal-source terms apply
  • OTIS Mock AIME 2024-2025#19 of 15898.9%

    +1.7 pts vs Gemini 3.7 Flash (predecessor)

    • high98.9%
    Source:Epoch AI evaluation
  • SimpleQA Verified#10 of 8769.7%

    +0.5 pts vs Gemini 3.7 Flash (predecessor)

    • high69.7%
    Source:Epoch AI evaluation

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Arena ratings: Arena (lmarena-ai/leaderboard-dataset), licensed CC BY 4.0. Ratings come from pairwise human votes and are only comparable within one leaderboard; the interval is Arena's 95% confidence interval. Model names are matched to the Vertical Tabs catalog.

To give you the best experience, we use cookies for personalization, insights, and performance. Learn more.