Reasoning
Like-for-like- GPT-5.4 Pro
- 83.3
- GPT-5.6 Luna
- 59.5
- Weighted basis
- 1 vs 1 rows
- Reading
- GPT-5.4 Pro leads
Model comparison
GPT-5.4 Pro vs GPT-5.6 Luna
Updated August 10, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
GPT-5.6 Luna has the higher public score estimate, 66.87 versus 60.08, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
6 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
Which one for your work
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Chat turn cost
1K fresh input + 500 output tokens
GPT-5.6 Luna
GPT-5.6 Luna has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
GPT-5.6 Luna
GPT-5.6 Luna has the lower estimated token cost for this stated workload. GPT-5.4 Pro has no published cached-input rate, so cached tokens use its listed input rate.
Confidence: rate-fallback
Show secondary and unsupported calls
Repository review cost
50K fresh input + 3K output tokens
GPT-5.6 Luna
GPT-5.6 Luna has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
No shared weighted benchmark basis supports a winner.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Long documents
Prompts that approach the documented context limit
No clear pick
The documented context windows are equal.
Confidence: documented
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
5 GPT-5.4 Pro only
6 Shared
16 GPT-5.6 Luna only
- Shared results
- 6
- GPT-5.4 Pro only
- 5
- GPT-5.6 Luna only
- 16
- Like-for-like categories
- 3 / 8
1 category uses different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Math
Like-for-like- GPT-5.4 Pro
- 46.9
- GPT-5.6 Luna
- 73.6
- Weighted basis
- 2 vs 2 rows
- Reading
- GPT-5.6 Luna leads
Multimodal
Like-for-like- GPT-5.4 Pro
- 94.0
- GPT-5.6 Luna
- 78.4
- Weighted basis
- 1 vs 1 rows
- Reading
- GPT-5.4 Pro leads
Agentic
Directional only- GPT-5.4 Pro
- 89.3
- GPT-5.6 Luna
- 84.1
- Weighted basis
- 1 vs 2 rows
- Reading
- Directional only
Coding
Not comparable- GPT-5.4 Pro
- Not measured
- GPT-5.6 Luna
- 62.7
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Knowledge
Not comparable- GPT-5.4 Pro
- 58.7
- GPT-5.6 Luna
- 92.3
- Weighted basis
- 1 vs 1 rows
- Reading
- Not comparable
Multilingual
Not comparable- GPT-5.4 Pro
- Not measured
- GPT-5.6 Luna
- Not measured
- Weighted basis
- 0 vs 0 rows
- Reading
- Not comparable
Instruction following
Not comparable- GPT-5.4 Pro
- Not measured
- GPT-5.6 Luna
- Not measured
- Weighted basis
- 0 vs 0 rows
- Reading
- Not comparable
| Category | GPT-5.4 Pro | GPT-5.6 Luna | Weighted basis | Reading |
|---|---|---|---|---|
| Reasoning | 83.3 | 59.5 | Like-for-like1 vs 1 rows | GPT-5.4 Pro leads |
| Math | 46.9 | 73.6 | Like-for-like2 vs 2 rows | GPT-5.6 Luna leads |
| Multimodal | 94.0 | 78.4 | Like-for-like1 vs 1 rows | GPT-5.4 Pro leads |
| Agentic | 89.3 | 84.1 | Directional only1 vs 2 rows | Directional only |
| Coding | Not measured | 62.7 | Not comparable0 vs 1 rows | Not comparable |
| Knowledge | 58.7 | 92.3 | Not comparable1 vs 1 rows | Not comparable |
| Multilingual | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Instruction following | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
Shape of the matched evidence
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
FrontierMath v2 (Tiers 1-3)
Math
ARC-AGI-2
Reasoning
FrontierMath v2 (Tier 4)
Math
MMMU-Pro
Multimodal
BrowseComp
Agentic
What each workload costs
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
Chat turn
1K fresh input + 500 output tokens
- GPT-5.4 Pro
- $0.12
- Fits in one request
- GPT-5.6 Luna
- $0.0008
- Fits in one request
GPT-5.6 Luna has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- GPT-5.4 Pro
- $2.04
- Fits in one request
- GPT-5.6 Luna
- $0.0136
- Fits in one request
GPT-5.6 Luna has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- GPT-5.4 Pro
- $8.40
- Fits in one request
- Cached input priced at the published list-input rate
- GPT-5.6 Luna
- $0.02
- Fits in one request
GPT-5.6 Luna has the lower modeled cost
GPT-5.4 Pro has no published cached-input rate, so cached tokens use its listed input rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationGPT-5.4 ProGPT-5.6 Luna
Context window
Maximum documented context; output-token limits may be lower.
GPT-5.4 Pro
GPT-5.6 Luna
1.05M
OpenAI model catalogAPI model ID
GPT-5.4 Pro
gpt-5.4-pro
OpenAI GPT-5.4 Pro model documentationGPT-5.6 Luna
gpt-5.6-luna
OpenAI model catalogCached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.4 Pro
Not published
OpenAI pricingGPT-5.6 Luna
$0.02 per 1M cached input tokens
OpenAI pricingDocumented inputs
GPT-5.4 Pro
text, image
OpenAI model catalogGPT-5.6 Luna
text, image
OpenAI model catalogDocumented outputs
GPT-5.4 Pro
GPT-5.6 Luna
Provider availability
GPT-5.4 Pro
Generally Available · OpenAI Responses API
OpenAI model catalogGPT-5.6 Luna
Generally Available · OpenAI Responses API
OpenAI model catalogReasoning profile
GPT-5.4 Pro
Reasoning
GPT-5.6 Luna
Reasoning
Weight access
GPT-5.4 Pro
Proprietary
GPT-5.6 Luna
Proprietary
License
GPT-5.4 Pro
Proprietary
GPT-5.6 Luna
Proprietary
Release date
GPT-5.4 Pro
2026-03-05
GPT-5.6 Luna
2026-07-09
If you already use one of these models
- Deployment change
- Both entries list OpenAI as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
- Quality signal
- GPT-5.6 Luna has the higher public score estimate, 66.87 versus 60.08, but the 90% score intervals overlap.
- Workload cost
- Repository review: $2.04 vs $0.0136. Cache-heavy agent loop: $8.40 vs $0.02.
- Context tradeoff
- Both models list 1.05M.
Run the same representative tasks against both endpoints before changing production traffic.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence27 rows
Agentic
BrowseComp
GPT-5.4 Pro89.3%SourceGPT-5.6 Luna83.3%SourceGPT-5.4 Pro leads this result
Terminal-Bench 2.0
GPT-5.4 Pro—GPT-5.6 Luna84.7%SourceNot directly comparable
OSWorld 2.0
GPT-5.4 Pro—GPT-5.6 Luna45.6%SourceNot directly comparable
CyberGym
GPT-5.4 Pro—GPT-5.6 Luna77.9%SourceNot directly comparable
ExploitGym
GPT-5.4 Pro—GPT-5.6 Luna12.4%SourceNot directly comparable
Toolathlon
GPT-5.4 Pro—GPT-5.6 Luna53.4%SourceNot directly comparable
Coding
SWE-bench Pro
GPT-5.4 Pro—GPT-5.6 Luna62.7%SourceNot directly comparable
Terminal-Bench 2.0
GPT-5.4 Pro—GPT-5.6 Luna84.7%SourceNot directly comparable
deepSwe
GPT-5.4 Pro—GPT-5.6 Luna67.2%SourceNot directly comparable
FrontierCode 1.1 Extended
GPT-5.4 Pro—GPT-5.6 Luna55.1%SourceNot directly comparable
cursorBench32
GPT-5.4 Pro—GPT-5.6 Luna61.1%SourceNot directly comparable
Reasoning
Knowledge
HLE
GPT-5.4 Pro58.7%SourceGPT-5.6 Luna—Not directly comparable
FrontierScience
GPT-5.4 Pro36.7%SourceGPT-5.6 Luna—Not directly comparable
FrontierScience Research
GPT-5.4 Pro36.7%SourceGPT-5.6 Luna—Not directly comparable
HLE w/o tools
GPT-5.4 Pro42.7%SourceGPT-5.6 Luna—Not directly comparable
GPQA
GPT-5.4 Pro—GPT-5.6 Luna92.3%SourceNot directly comparable
GPQA-D
GPT-5.4 Pro—GPT-5.6 Luna92.3%SourceNot directly comparable
HealthBench Professional
GPT-5.4 Pro—GPT-5.6 Luna55.7%SourceNot directly comparable
HealthBench Hard
GPT-5.4 Pro—GPT-5.6 Luna32.0%SourceNot directly comparable
Math
IPhO 2025 (Theory)
GPT-5.4 Pro93.5%SourceGPT-5.6 Luna—Not directly comparable
FrontierMath (legacy)
GPT-5.4 Pro50%SourceGPT-5.6 Luna78.6%SourceGPT-5.6 Luna leads this result
FrontierMath v2 (Tiers 1-3)
GPT-5.4 Pro50.000%SourceGPT-5.6 Luna78.600%SourceGPT-5.6 Luna leads this result
FrontierMath v2 (Tier 4)
GPT-5.4 Pro37.500%SourceGPT-5.6 Luna58.500%SourceGPT-5.6 Luna leads this result
Multimodal
Frequently asked questions
Which is better, GPT-5.4 Pro or GPT-5.6 Luna?
GPT-5.6 Luna has the higher public score estimate, 66.87 versus 60.08, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Which is better for coding, GPT-5.4 Pro or GPT-5.6 Luna?
The published evidence does not provide a shared weighted coding basis for both models, so BenchLM does not name a coding winner.
Which is better for agentic tasks, GPT-5.4 Pro or GPT-5.6 Luna?
The current agentic tasks averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which costs less, GPT-5.4 Pro or GPT-5.6 Luna?
For the stated presets, chat costs $0.12 on GPT-5.4 Pro and $0.0008 on GPT-5.6 Luna; repository review costs $2.04 and $0.0136; the cache-heavy agent loop costs $8.40 and $0.02. GPT-5.4 Pro has no published cached-input rate, so cached tokens use its listed input rate.
Which has the larger context window, GPT-5.4 Pro or GPT-5.6 Luna?
Both models list the same context window, 1.05M.
Related comparisons
Last updated August 10, 2026
Watch GPT-5.4 Pro vs GPT-5.6 Luna
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.