Model comparison
GPT-5.3 Codex vs GPT-5.6 Luna
Head-to-head evidence from 12 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: GPT-5.3 Codex #26 (Supported); GPT-5.6 Luna #22 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. GPT-5.3 Codex and GPT-5.6 Luna share 12 comparable benchmark results. 2 of 8 categories are comparable. 9 results are unique to GPT-5.3 Codex; 29 to GPT-5.6 Luna.
Updated July 23, 2026- Shared results
- 12
- GPT-5.3 Codex only
- 9
- GPT-5.6 Luna only
- 29
- Comparable categories
- 2 / 8
Pick GPT-5.6 Luna if you want the stronger benchmark profile. GPT-5.3 Codex only becomes the better choice if coding is the priority.
Confidence note. This is a partial-evidence comparison with 12 shared benchmark results across 5 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
GPT-5.6 Luna has the cleaner BenchAlign overall profile here, landing at 67.17 versus 66.69. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
GPT-5.6 Luna's sharpest advantage is in agentic, where it averages 84.1 against 71.4. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 77.3% to 84.7%. GPT-5.3 Codex does hit back in coding, so the answer changes if that is the part of the workload you care about most.
GPT-5.3 Codex is also the more expensive model on tokens at $1.75 input / $14.00 output per 1M tokens, versus $1.00 input / $6.00 output per 1M tokens for GPT-5.6 Luna. That is roughly 2.3x on output cost alone. GPT-5.6 Luna gives you the larger context window at 1M, compared with 400K for GPT-5.3 Codex.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | GPT-5.3 Codex | Δ | GPT-5.6 Luna |
|---|---|---|---|
| Agentic | GPT-5.3 Codex71.4 | Margin→ 12.7 | GPT-5.6 Luna84.1 |
| Coding | GPT-5.3 Codex67.2 | Margin← 4.5 | GPT-5.6 Luna62.7 |
| Knowledge | GPT-5.3 CodexNot measured | MarginNo overlap | GPT-5.6 Luna92.3 |
| Math | GPT-5.3 CodexNot measured | MarginNo overlap | GPT-5.6 Luna73.6 |
| Multimodal | GPT-5.3 CodexNot measured | MarginNo overlap | GPT-5.6 Luna78.4 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 77.3%B 84.7%Winner: GPT-5.6 LunaΔ 7.4Terminal-Bench 2.0: GPT-5.3 Codex scored 77.3%; GPT-5.6 Luna scored 84.7%. GPT-5.6 Luna wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 56.8%B 62.7%Winner: GPT-5.6 LunaΔ 5.9SWE-bench Pro: GPT-5.3 Codex scored 56.8%; GPT-5.6 Luna scored 62.7%. GPT-5.6 Luna wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | GPT-5.3 Codex | GPT-5.6 Luna | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | GPT-5.3 Codex$1.75 input / $14 output | GPT-5.6 Luna$1 input / $6 output | GPT-5.6 Luna has the lower combined listed price. |
| Generation speedtokens per second | GPT-5.3 Codex79 tok/s | GPT-5.6 LunaNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | GPT-5.3 Codex88.26 s | GPT-5.6 LunaNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | GPT-5.3 Codex400K | GPT-5.6 Luna1M | GPT-5.6 Luna lists the larger context window. |
Benchmark Deep Dive
AgenticGPT-5.6 Luna wins19 benchmarks
| Benchmark | GPT-5.3 Codex | GPT-5.6 Luna | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 77.3% | 84.7% | GPT-5.6 Luna leads |
| OSWorld-VerifiedSource | 64.7% | — | Not comparable |
| τ²-bench resultsSource | 86% | — | Not comparable |
| Gert LabsSource | 57.47% | — | Not comparable |
| JobBenchSource | 33.7% | — | Not comparable |
| BrowseCompSource | — | 83.3% | Not comparable |
| OSWorld 2.0Source | — | 45.6% | Not comparable |
| CyberGymSource | — | 77.9% | Not comparable |
| ExploitGymSource | — | 12.4% | Not comparable |
| ToolathlonSource | — | 53.4% | Not comparable |
| AA Agentic IndexSource | — | 45.6% | Not comparable |
| GDPval-AASource | — | 54.2% | Not comparable |
| GDPval-AASource | — | 1584 | Not comparable |
| AA Harvey LABSource | — | 87.9% | Not comparable |
| AA ITBenchSource | — | 40.3% | Not comparable |
| AA Tau3 BankingSource | — | 27.2% | Not comparable |
| AA AutomationBenchSource | — | 42.2% | Not comparable |
| aaTerminalBench21Source | — | 80.9% | Not comparable |
| APEX-Agents-AASource | — | 35.8% | Not comparable |
CodingGPT-5.3 Codex wins10 benchmarks
| Benchmark | GPT-5.3 Codex | GPT-5.6 Luna | Result |
|---|---|---|---|
| SWE-bench VerifiedSource | 85% | — | Not comparable |
| SWE-bench ProSource | 56.8% | 62.7% | GPT-5.6 Luna leads |
| SWE-RebenchSource | 58.2% | — | Not comparable |
| Vibe Code BenchSource | 61.77% | — | Not comparable |
| AA-SciCodeSource | 53.2% | 52.5% | GPT-5.3 Codex leads |
| Terminal-Bench 2.0Source | — | 84.7% | Not comparable |
| deepSweSource | — | 67.2% | Not comparable |
| FrontierCode 1.1 ExtendedSource | — | 55.1% | Not comparable |
| cursorBench32Source | — | 61.1% | Not comparable |
| AA Coding IndexSource | — | 71.5% | Not comparable |
Reasoning3 benchmarks
Knowledge10 benchmarks
| Benchmark | GPT-5.3 Codex | GPT-5.6 Luna | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 44.3% | 51.2% | GPT-5.6 Luna leads |
| AA-GPQA DiamondSource | 91.5% | 91.1% | GPT-5.3 Codex leads |
| AA-HLESource | 39.9% | 37.2% | GPT-5.3 Codex leads |
| AA-Omniscience IndexSource | 9.9% | -11.2% | GPT-5.3 Codex leads |
| AA-Omniscience AccuracySource | 51.8% | 41.5% | GPT-5.3 Codex leads |
| AA-Omniscience Hallucination RateSource | 86.9% | 90.1% | GPT-5.3 Codex leads |
| GPQASource | — | 92.3% | Not comparable |
| GPQA-DSource | — | 92.3% | Not comparable |
| HealthBench ProfessionalSource | — | 55.7% | Not comparable |
| HealthBench HardSource | — | 32.0% | Not comparable |
Math3 benchmarks
Multimodal4 benchmarks
Inst. Following1 benchmarks
| Benchmark | GPT-5.3 Codex | GPT-5.6 Luna | Result |
|---|---|---|---|
| AA-IFBenchSource | 75.4% | — | Not comparable |
Frequently Asked Questions (3)
Which is better, GPT-5.3 Codex or GPT-5.6 Luna?
GPT-5.6 Luna is ahead on BenchLM's BenchAlign leaderboard, 67.17 to 66.69. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 77.3% and 84.7%.
Which is better for coding, GPT-5.3 Codex or GPT-5.6 Luna?
GPT-5.3 Codex has the edge for coding in this comparison, averaging 67.2 versus 62.7. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, GPT-5.3 Codex or GPT-5.6 Luna?
GPT-5.6 Luna has the edge for agentic tasks in this comparison, averaging 84.1 versus 71.4. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.