Coding
Like-for-like- GPT-5.6 Luna
- 62.7
- Hy4 preview
- 65.7
- Weighted basis
- 1 vs 1 rows
- Reading
- Hy4 preview leads
Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.
Start the free Radar BriefUpdated August 28, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
Hy4 preview has the higher public score estimate, 79.16 versus 67.35, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
5 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
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.
Code generation, repair, and software-engineering tasks
Hy4 preview
Hy4 preview leads on the same 1 weighted benchmark row.
Confidence: limited
Prompts that approach the documented context limit
GPT-5.6 Luna
GPT-5.6 Luna has the larger documented context window.
Confidence: documented
Tool use, computer use, and multi-step task completion
Not enough matched evidence
No shared weighted benchmark basis supports a winner.
Confidence: limited
1K fresh input + 500 output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
50K fresh input + 3K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
1 category uses different evidence sets. Those rows remain visible for coverage context but do not name a winner.
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.
| Category | GPT-5.6 Luna | Hy4 preview | Weighted basis | Reading |
|---|---|---|---|---|
| Coding | 62.7 | 65.7 | Like-for-like1 vs 1 rows | Hy4 preview leads |
| Knowledge | 92.3 | 60.4 | Directional only1 vs 2 rows | Directional only |
| Agentic | 84.1 | Not measured | Not comparable2 vs 0 rows | Not comparable |
| Reasoning | 59.5 | Not measured | Not comparable1 vs 0 rows | Not comparable |
| Math | 73.6 | Not measured | Not comparable2 vs 0 rows | Not comparable |
| Multilingual | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Multimodal | 78.4 | 66.2 | Not comparable1 vs 1 rows | Not comparable |
| Instruction following | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
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.
SWE-bench Pro
Coding
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.
1K fresh input + 500 output tokens
Hy4 preview has no comparable published API token rate.
50K fresh input + 3K output tokens
Hy4 preview has no comparable published API token rate.
200K cached + 20K fresh input + 10K output tokens
Hy4 preview has no comparable published API token rate.
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
Maximum documented context; output-token limits may be lower.
GPT-5.6 Luna
1.05M
OpenAI model catalogHy4 preview
GPT-5.6 Luna
gpt-5.6-luna
OpenAI model catalogHy4 preview
tencent/Hy4-preview
Tencent Hy4 preview model cardA missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.6 Luna
$0.1 per 1M cached input tokens
OpenAI API pricingHy4 preview
No comparable hosted API rate
Tencent Hy4 preview model cardGPT-5.6 Luna
text, image
OpenAI model catalogHy4 preview
Not sourced
GPT-5.6 Luna
Hy4 preview
Not sourced
GPT-5.6 Luna
Generally Available · OpenAI Responses API
OpenAI model catalogHy4 preview
Not sourced
GPT-5.6 Luna
Reasoning
Hy4 preview
Reasoning
GPT-5.6 Luna
Proprietary
Hy4 preview
Open Weight
GPT-5.6 Luna
Proprietary
Hy4 preview
Open Weight
GPT-5.6 Luna
2026-07-09
Hy4 preview
2026-08-28
Run the same representative tasks against both endpoints before changing production traffic.
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Terminal-Bench 3.0
Not directly comparable
Terminal-Bench 2.0
Not directly comparable
BrowseComp
Not directly comparable
OSWorld 2.0
Not directly comparable
CyberGym
Hy4 preview leads this result
ExploitGym
Not directly comparable
Toolathlon
Not directly comparable
Terminal-Bench 2.1
Not directly comparable
WideResearch
Not directly comparable
DRACO
Not directly comparable
MCP Atlas
Not directly comparable
Toolathlon-Verified
Not directly comparable
APEX-Agents
Not directly comparable
skillsBench
Not directly comparable
JobBench
Not directly comparable
Agents' Last Exam
Not directly comparable
AutomationBench
Not directly comparable
BankerToolBench
Not directly comparable
HLE w/ tools
Not directly comparable
SWE-bench Pro
Hy4 preview leads this result
Terminal-Bench 2.0
Not directly comparable
deepSwe
GPT-5.6 Luna leads this result
FrontierCode 1.1 Extended
Not directly comparable
cursorBench32
Not directly comparable
VulcanBench v3
Not directly comparable
Terminal-Bench 2.1
Not directly comparable
SWE Multilingual
Not directly comparable
NL2Repo
Not directly comparable
ProgramBench
Not directly comparable
PostTrain Bench
Not directly comparable
sweMarathon
Not directly comparable
GPQA
Tie
GPQA-D
Tie
HealthBench Professional
Not directly comparable
HealthBench Hard
Not directly comparable
HLE
Not directly comparable
HLE w/o tools
Not directly comparable
FrontierMath (legacy)
Not directly comparable
FrontierMath v2 (Tiers 1-3)
Not directly comparable
FrontierMath v2 (Tier 4)
Not directly comparable
Apex
Not directly comparable
Hy4 preview has the higher public score estimate, 79.16 versus 67.35, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Hy4 preview leads the like-for-like coding comparison across 1 shared weighted benchmark row.
The published evidence does not provide a shared weighted agentic tasks basis for both models, so BenchLM does not name a agentic tasks winner.
Both models do not have comparable published API token rates, so this page does not name a universal price winner.
GPT-5.6 Luna has the larger documented context window: 1.05M, compared with 1M.
Last updated August 28, 2026
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.