Glimmer Index
Plus $4/mo

GPT-5.4

OpenAI · Released Mar 5, 2026 · Closed weights

Agent-ready GPT for coding and computer-use workflows at a lower cost

Strongest at
Science
Top 15% of the past year's models · 3 of 4 benchmarks
Price per 1M tokens
$2.5 / $15
input / output · OpenAI
Context window
1.05M
tokens in · 128K out
Since the last one
84 days
after GPT-5.2
Lab announcement ↗Open in the timeline

Capability profile

Scored against 91 models released in the past year. 50 is the average model; further out is better.

This modelGPT-5.2Median of the past year
5090Median of the past yearCoding: 52 · Better than 34%Writing: 63 · Top 20%Reading & documents: 63 · Top 21%Reasoning: 68 · Top 27%Math: 64 · Top 22%Science: 71 · Top 15%Visual reasoning: 54 · Better than 42%Computer use: 65 · Better than 44%GPT-5.2: 57GPT-5.2: 59GPT-5.2: 59GPT-5.2: 54GPT-5.2: 66CodingWritingReadingReasoningMathScienceVisionComputer use
Coding52−5.2Better than 34%
Writing63+3.6Top 20%
Reading & documents63Top 21% · 1 test
Reasoning68+8.9Top 27%
Math64+9.9Top 22%
Science71+4.9Top 15%
Visual reasoning54Better than 42% · 1 test
Computer use65Better than 44%
Where these scores come from

Each capability combines the published benchmarks listed here, rescaled so 50 is the reference average and 15 points is one standard deviation. A hollow dot rests on a single benchmark.

Price against capability

The shaded corner holds anything better and cheaper. Prices blend input and output 3 to 1.

4 models from the past year score higher on Science for less: Claude Sonnet 5.5, Gemini 3.1 Pro Preview, GPT-5.3 Codex and 1 more.

$0.1$0.3$1$3$10$30406080100GPT-6.1 Terra: 69, $2 per 1MDeepSeek V4.1 Flash 1017: 69, $0.52 per 1MMuse Spark Open: 66, $2 per 1MClaude Haiku 5.5: 65, $0.2 per 1MMistral Large 4: 65, $2.06 per 1MGemini 4 Argon: 71, $4 per 1MGPT-6.1 Sol: 70, $4 per 1MClaude Sonnet 5.5: 76, $4 per 1MGPT-6 Sol: 69, $4 per 1MGPT-6 Luna: 67, $0.2 per 1MClaude Opus 5.5: 108, $8 per 1MGrok 4.7: 61, $3 per 1MDeepSeek V4.1 Flash: 62, $0.21 per 1MGPT-6 Astra: 101, $20 per 1MGemini 3.8 Flash: 70, $1.5 per 1MClaude Fable 5.1: 90, $20 per 1MGLM-5.3-Flash: 67, $0.24 per 1MGLM-5.3: 60, $2.15 per 1MGemini 3.7 Flash: 54, $1.5 per 1MDeepSeek V4 Pro 0813: 66, $0.99 per 1MQwen3.8 2.4T A95B: 70, $3 per 1MGrok 4.6: 58, $3 per 1MMuse Spark 1.2: 60, $2 per 1MQwen3.8 Max: 68, $3 per 1MDeepSeek V4 Flash 0731: 64, $0.15 per 1MClaude Opus 5: 79, $10 per 1MGemini 3.6 Flash: 63, $1.5 per 1MGemini 3.5 Flash Lite: 53, $0.85 per 1MKimi K3: 62, $6 per 1MQwen3.7 Flash: 62, $0.055 per 1MGPT-5.6 Terra: 61, $4.5 per 1MGPT-5.6 Sol: 73, $8 per 1MGPT-5.6 Luna: 54, $0.45 per 1MGrok 4.5: 59, $3 per 1MClaude Sonnet 5: 66, $4 per 1MGLM-5.2: 67, $2.15 per 1MKimi K2.7 Code: 60, $1.71 per 1MClaude Fable 5: 71, $20 per 1MQwen3.7 Plus: 65, $0.7 per 1MClaude Opus 4.8: 63, $10 per 1MQwen3.7 Max: 67, $3.75 per 1MGemini 3.5 Flash: 65, $3.38 per 1MGPT-5.5 Instant: 62, $11.2 per 1MDeepSeek V4 Pro: 51, $1.19 per 1MDeepSeek V4 Flash: 49, $0.18 per 1MGPT-5.5 Pro: 69, $67.5 per 1MGPT-5.5: 73, $11.2 per 1MQwen3.6 27B: 64, $1.35 per 1MKimi K2.6: 62, $1.71 per 1MGrok 4.3: 52, $1.56 per 1MClaude Opus 4.7: 70, $10 per 1MQwen3.6 35B-A3B: 52, $0.84 per 1MGLM-5.1: 62, $2.15 per 1MGemma 4 31B IT: 57, $0.25 per 1MGPT-5.4 nano: 56, $0.46 per 1MGPT-5.4 mini: 62, $1.69 per 1MNemotron 3 Super 120B A12B: 44, $0.82 per 1MGPT-5.4 Pro: 75, $67.5 per 1MGemini 3.1 Flash Lite Preview: 54, $0.56 per 1MGemini 3.1 Flash Lite: 50, $0.56 per 1MQwen3.5 Flash: 62, $0.18 per 1MQwen3.5 9B: 59, $0.19 per 1MQwen3.5 35B-A3B: 62, $0.69 per 1MQwen3.5 27B: 45, $0.82 per 1MGemini 3.1 Pro Preview: 73, $4.5 per 1MClaude Sonnet 4.6: 63, $6 per 1MQwen3.5 Plus: 63, $0.9 per 1MQwen3.5 397B-A17B: 64, $1.35 per 1MGLM-5: 58, $1.55 per 1MGPT-5.3 Codex: 72, $4.81 per 1MClaude Opus 4.6: 69, $10 per 1MQwen3 Coder Next: 42, $0.29 per 1MKimi K2.5: 53, $1 per 1MGLM-4.7: 62, $1 per 1MGemini 3 Flash Preview: 63, $1.12 per 1MDeepSeek V3.2: 62, $0.32 per 1MClaude Opus 4.5: 60, $10 per 1MGPT-5.1: 61, $3.44 per 1MKimi K2 Thinking: 47, $1.07 per 1MGPT-5.2: 66, $4.81 per 1MGPT-5.4: 71, $5.62 per 1MGPT-5.4Price per 1M tokens →↑ Science$0.1$1$10406080100GPT-6.1 Terra: 69, $2 per 1MDeepSeek V4.1 Flash 1017: 69, $0.52 per 1MMuse Spark Open: 66, $2 per 1MClaude Haiku 5.5: 65, $0.2 per 1MMistral Large 4: 65, $2.06 per 1MGemini 4 Argon: 71, $4 per 1MGPT-6.1 Sol: 70, $4 per 1MClaude Sonnet 5.5: 76, $4 per 1MGPT-6 Sol: 69, $4 per 1MGPT-6 Luna: 67, $0.2 per 1MClaude Opus 5.5: 108, $8 per 1MGrok 4.7: 61, $3 per 1MDeepSeek V4.1 Flash: 62, $0.21 per 1MGPT-6 Astra: 101, $20 per 1MGemini 3.8 Flash: 70, $1.5 per 1MClaude Fable 5.1: 90, $20 per 1MGLM-5.3-Flash: 67, $0.24 per 1MGLM-5.3: 60, $2.15 per 1MGemini 3.7 Flash: 54, $1.5 per 1MDeepSeek V4 Pro 0813: 66, $0.99 per 1MQwen3.8 2.4T A95B: 70, $3 per 1MGrok 4.6: 58, $3 per 1MMuse Spark 1.2: 60, $2 per 1MQwen3.8 Max: 68, $3 per 1MDeepSeek V4 Flash 0731: 64, $0.15 per 1MClaude Opus 5: 79, $10 per 1MGemini 3.6 Flash: 63, $1.5 per 1MGemini 3.5 Flash Lite: 53, $0.85 per 1MKimi K3: 62, $6 per 1MQwen3.7 Flash: 62, $0.055 per 1MGPT-5.6 Terra: 61, $4.5 per 1MGPT-5.6 Sol: 73, $8 per 1MGPT-5.6 Luna: 54, $0.45 per 1MGrok 4.5: 59, $3 per 1MClaude Sonnet 5: 66, $4 per 1MGLM-5.2: 67, $2.15 per 1MKimi K2.7 Code: 60, $1.71 per 1MClaude Fable 5: 71, $20 per 1MQwen3.7 Plus: 65, $0.7 per 1MClaude Opus 4.8: 63, $10 per 1MQwen3.7 Max: 67, $3.75 per 1MGemini 3.5 Flash: 65, $3.38 per 1MGPT-5.5 Instant: 62, $11.2 per 1MDeepSeek V4 Pro: 51, $1.19 per 1MDeepSeek V4 Flash: 49, $0.18 per 1MGPT-5.5 Pro: 69, $67.5 per 1MGPT-5.5: 73, $11.2 per 1MQwen3.6 27B: 64, $1.35 per 1MKimi K2.6: 62, $1.71 per 1MGrok 4.3: 52, $1.56 per 1MClaude Opus 4.7: 70, $10 per 1MQwen3.6 35B-A3B: 52, $0.84 per 1MGLM-5.1: 62, $2.15 per 1MGemma 4 31B IT: 57, $0.25 per 1MGPT-5.4 nano: 56, $0.46 per 1MGPT-5.4 mini: 62, $1.69 per 1MNemotron 3 Super 120B A12B: 44, $0.82 per 1MGPT-5.4 Pro: 75, $67.5 per 1MGemini 3.1 Flash Lite Preview: 54, $0.56 per 1MGemini 3.1 Flash Lite: 50, $0.56 per 1MQwen3.5 Flash: 62, $0.18 per 1MQwen3.5 9B: 59, $0.19 per 1MQwen3.5 35B-A3B: 62, $0.69 per 1MQwen3.5 27B: 45, $0.82 per 1MGemini 3.1 Pro Preview: 73, $4.5 per 1MClaude Sonnet 4.6: 63, $6 per 1MQwen3.5 Plus: 63, $0.9 per 1MQwen3.5 397B-A17B: 64, $1.35 per 1MGLM-5: 58, $1.55 per 1MGPT-5.3 Codex: 72, $4.81 per 1MClaude Opus 4.6: 69, $10 per 1MQwen3 Coder Next: 42, $0.29 per 1MKimi K2.5: 53, $1 per 1MGLM-4.7: 62, $1 per 1MGemini 3 Flash Preview: 63, $1.12 per 1MDeepSeek V3.2: 62, $0.32 per 1MClaude Opus 4.5: 60, $10 per 1MGPT-5.1: 61, $3.44 per 1MKimi K2 Thinking: 47, $1.07 per 1MGPT-5.2: 66, $4.81 per 1MGPT-5.4: 71, $5.62 per 1MGPT-5.4Price per 1M tokens →↑ Science

OpenAI $2.5 / $15 · OpenRouter $2.5 / $15 · Cached input $0.25

Since GPT-5.2

8 releases in the GPT-5.5 line.

GPT-4oMay 13, 2024336dGPT-4.1Apr 14, 2025115dGPT-5Aug 7, 202598dGPT-5.1Nov 13, 202528dGPT-5.2Dec 11, 202584dGPT-5.4Mar 5, 202649dGPT-5.5Apr 23, 2026
Released84 days later
Price in / out$2.5 / $15+17%
Context1.05M2.6×
Coding52−5.2
Writing63+3.6
Reasoning68+8.9
Math64+9.9
Science71+4.9

A newer release is out: GPT-5.5.

ECI STANDING#22 of 270
SCORE157.31155.38–159.8 CI
FRONTIER LAGLaunched in the leadNothing released before it had scored higher.
SCORES LIKE THESE MODELSGemini 3.8 FlashGoogle DeepMind · Sep 2026157.44Qwen 3.8 MaxAlibaba · Aug 2026157.4Muse Spark 1.3Meta AI · Sep 2026157.21GPT-5.3 CodexOpenAI · Feb 2026157.09
Full Epoch Capabilities Index ranking ↗

Epoch benchmark boards

Best result on each board it has run. The tick marks the leader.

Terminal Bench
82%#2 of 50
CL-bench
28%#1 of 21
GPQA diamond
93%#19 of 234
WeirdML
78%#15 of 141
CL-bench Life
22%#2 of 14
OTIS Mock AIME 2024-2025
98%#32 of 216
DTBench
94%#28 of 179
METR Time Horizons
74%#7 of 42
16 more boards
ARC-AGI-2
74%#19 of 103
ARC-AGI
94%#21 of 107
Mystery Game Puzzles
37%#19 of 88
FrontierMath-Tiers-1-3-v2-Private
79%#20 of 90
Chess Puzzles
48%#21 of 93
SWE-Bench verified
77%#8 of 32
FrontierMath-Tier-4-v2-Private
49%#22 of 72
ProofBench
56%#25 of 81
GSO-Bench
31%#12 of 35
APEX-Agents
52%#24 of 51
SimpleQA Verified
45%#41 of 85
Furniture Assembly
38%#20 of 35
EBR-bench
25%#16 of 28
MirrorCode
16%#8 of 10
DeepResearch Bench
35%#25 of 26
PostTrainBench
19%#11 of 11

OpenAI's past year

24 release days in the past year, about one every 15 days. The next is expected around Nov 8, 2026. Forecasts ↗

NovDecJanFebMarAprMayJunJulAugSepOctNov

Specifications

RELEASE DATEMar 5, 2026
WEIGHTSClosed weights
CONTEXT WINDOW1.05M
MAX OUTPUT128K
REASONINGNot listed
TOOL CALLINGNot listed
INPUTStext, image/jpeg, image/png, image/gif, image/webp, application/pdf
OUTPUTStext

Models.dev record ↗

Specifications and prices from Models.dev; availability and limits may differ by provider. Benchmarks from Epoch AI (CC BY 4.0), LiveBench and the other sources named above.