vs
8:5
overall capability wins (2 ties across 15 scored). GPT-5.4 leads overall.
- Reasoning4:2
- Coding2:3
- Arena1:0
- Tool use1:0
- Pricing0:3
- Speed0:2
Verdict
GPT-5.6 Luna leads coding (widest gap: +132.2 on WebDev Elo); GPT-5.4 edges reasoning & knowledge; GPT-5.4 ranks higher on LMArena (+13 Elo).
- GPT-5.6 Luna leads coding (widest gap: +132.2 on WebDev Elo)
- GPT-5.4 edges reasoning & knowledge
- GPT-5.4 ranks higher on LMArena (+13 Elo)
- GPT-5.6 Luna is ~12.5x cheaper on a blended token basis than GPT-5.4
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- GPT-5.4
- $6.25
- GPT-5.6 Luna
- $0.50
GPT-5.6 Luna is about 12.5x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
OpenAI
OpenAI
Family
GPT-5.x
GPT-5.x
License
Proprietary
Proprietary
Open weights
No
No
Release
Mar 5, 2026
Jul 9, 2026
Knowledge cutoff
-
-
API / provider
OpenAI
OpenAI
Modalities
text, image → text
text, image → text
Specs
Context window
1.1M
1.1M
Max output
128K
128K
Parameters
—
—
Pricing
Input $/1M
$2.50
$0.20
Output $/1M
$15
$1.20
Blended $/1M (3∶1)
$5.63
$0.45
Speed
tok/s
85
195
TTFT (s)
0.4
0.2
Reasoning
MMLU-Pro
—
—
GPQA Diamond
92.0
91.1
Humanity's Last Exam
43.7
39.5
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
50.8
42.7
AA-LCR v1.1
82.0
83.7
CritPt
23.4
20.6
MMMU-Pro
78.4
78.6
IFBench
73.9
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
57.7
62.7
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
78.3
80.9
Aider Polyglot
—
—
Terminal-Bench 3
—
14.3
BigCodeBench
—
—
SciCode
56.6
53.6
CursorBench
—
61.1
SWE-Rebench
57.6
43.6
NL2Repo-Bench
—
—
DeepSWE
—
67.0
WebDev Arena
1,387
1,519
Terminal-Bench 4.0
—
17.3
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,466
1,453
| Metric | GPT-5.4 | GPT-5.6 Luna | Delta |
|---|---|---|---|
| Identity | |||
| Organization | OpenAI | OpenAI | — |
| Family | GPT-5.x | GPT-5.x | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | Mar 5, 2026 | Jul 9, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | OpenAI | OpenAI | — |
| Modalities | text, image → text | text, image → text | — |
| Specs | |||
| Context window | 1.1M | 1.1M | tie |
| Max output | 128K | 128K | tie |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $2.50 | $0.20 | B +$2.30 |
| Output $/1M | $15 | $1.20 | B +$13.80 |
| Blended $/1M (3∶1) | $5.63 | $0.45 | B +$5.17 |
| Speed | |||
| tok/s | 85 | 195 | B +110 |
| TTFT (s) | 0.4 | 0.2 | B +0.20 |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 92.0 | 91.1 | A +0.9 pts |
| Humanity's Last Exam | 43.7 | 39.5 | A +4.2 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 50.8 | 42.7 | A +8.1 pts |
| AA-LCR v1.1 | 82.0 | 83.7 | B +1.7 pts |
| CritPt | 23.4 | 20.6 | A +2.8 pts |
| MMMU-Pro | 78.4 | 78.6 | B +0.2 pts |
| IFBench | 73.9 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | 57.7 | 62.7 | B +5.0 pts |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 78.3 | 80.9 | B +2.6 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | 14.3 | — |
| BigCodeBench | — | — | — |
| SciCode | 56.6 | 53.6 | A +3.0 pts |
| CursorBench | — | 61.1 | — |
| SWE-Rebench | 57.6 | 43.6 | A +14.0 pts |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | 67.0 | — |
| WebDev Arena | 1,387 | 1,519 | B +132 Elo |
| Terminal-Bench 4.0 | — | 17.3 | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,466 | 1,453 | A +13 Elo |
GPT-5.4
- Context
- 1.1M / 128K out
- Parameters
- —
- Price
- $2.50 / $15
- Speed
- 85 tok/s · 0.4s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Solid mid-2026 OpenAI workhorse between GPT-5 and the 5.5/5.6 flagships.
GPT-5.6 Luna
- Context
- 1.1M / 128K out
- Parameters
- —
- Price
- $0.20 / $1.20
- Speed
- 195 tok/s · 0.2s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Fastest, cheapest GPT-5.6 tier for high-volume chat, classification, and light coding; July 30, 2026 API cut dropped Standard rates 80% to $0.20/$1.20.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.