OpenAI, released Mar 5, 2026

GPT-5.4 Proprice, context, benchmarks and release details

69% confidence 69 percent, Medium confidence, 4 of 7 expected sources in
61.3
SI Score
#58 of 142 ranked models
Input, per 1M tokens
$30.00models.devFirst-party hosted API; MIT models.dev transcription. Provider documentation: https://platform.openai.com/docs/models. Exact canonical endpoint; lowest short-context Standard USD token tier; cache/batch discounts excluded. Deprecated endpoints excluded; openai/gpt-5.4-pro Retrieved Oct 9, 2026 · MIT
Open source ↗
Output, per 1M tokens
$180.00models.devFirst-party hosted API; MIT models.dev transcription. Provider documentation: https://platform.openai.com/docs/models. Exact canonical endpoint; lowest short-context Standard USD token tier; cache/batch discounts excluded. Deprecated endpoints excluded; openai/gpt-5.4-pro Retrieved Oct 9, 2026 · MIT
Open source ↗
Context window
1.1Mmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Max output
128Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released
Mar 5, 2026OpenAI API changelogPublished source fact Retrieved Oct 9, 2026 · factual citation
Open source ↗

How the score breaks down

Coding (weight 40 percent) —
Math (weight 15 percent) 63.6
Preference (weight 15 percent) —
Reasoning (weight 30 percent) 80.8

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. 56 Qwen3.6 Plus 61.6
  2. 57 GLM-4.7 61.3
  3. 58 GPT-5.4 Pro 61.3
  4. 59 Nemotron 3 Ultra 550B A55B 61.2
  5. 60 DeepSeek V4 Flash 0731 61.0

Full leaderboard

Benchmark results

14 benchmarks, 15 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.

ARC-AGI-1reasoning 94.5%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] VerifiedPublished Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 94.5
ARC-AGI-1 (public eval)reasoning 98.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] gpt-5-4-pro-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 98.3
ARC-AGI-1 (semi-private)reasoning 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. [variant] gpt-5-4-pro-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 94.5
ARC-AGI-2reasoning 83.3%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] VerifiedPublished Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 83.3
ARC-AGI-2 (public eval)reasoning 92.2%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] gpt-5-4-pro-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 92.2
ARC-AGI-2 (semi-private)reasoning 83.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] gpt-5-4-pro-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 83.3
GPQA Diamondreasoning 94.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] xhighPublished Mar 20, 2026 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 94.6 2 settings
  • xhigh effort 94.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] xhighPublished Mar 20, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 2 94.4%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] Published Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗
About GPQA Diamond
Humanity's Last Examreasoning 42.7%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] no toolsPublished Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 42.7
Humanity's Last Exam (Scale AI)reasoning 44.3%Humanity’s Last ExamPotential contamination warning: This model was evaluated after the public release of HLE, allowing model builder access to the prompts and solutions. [variant] Published Mar 23, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 44.3
Humanity's Last Exam (with tools)reasoning 58.7%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] with toolsPublished Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 58.7
FrontierMath Tier 4math 38%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] Tier 4Published Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 38.0
FrontierMath Tier 4 (v2)math 58.5%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] xhighPublished Jun 13, 2026 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 58.5
FrontierMath Tiers 1–3math 50%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] Tier 1-3Published Apr 23, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 50.0
FrontierMath Tiers 1–3 (v2)math 82.5%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] xhighPublished Jun 13, 2026 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 82.5

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
Nomodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
License
not yet reported
Input modalities
text, imagemodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
First seen by SuperIndex
Oct 8, 2026
Coverage
55% of expected source weight

Reported (4)

  • ARC PrizeOct 8, 2026
  • Epoch AI BenchmarkingOct 8, 2026
  • Humanity’s Last ExamOct 8, 2026
  • Official model cards via models.devOct 8, 2026

Awaiting (3)

  • LiveBench13% of weight
  • LMArena / Arena25% of weight
  • Terminal-Bench6% 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