Updated November 9, 2026. Rank says GPT-5.6 Sol is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.
GPT-5.6 Sol has the higher public score estimate, 74.36 versus 69.18, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 26 results are shared. Category rows resting on Estimated evidence or 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
GPT-5.6 Sol
GPT-5.6 Sol has the higher public coding point estimate, 67.2 to 60.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Tool use, computer use, and multi-step task completion
GPT-5.6 Sol
GPT-5.6 Sol has the higher public agentic point estimate, 67.9 to 59.5, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Prompts that approach the documented context limit
GPT-5.6 Sol
GPT-5.6 Sol has the larger documented context window.
1K fresh input + 500 output tokens
GPT-5.6 Sol
GPT-5.6 Sol has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
GPT-5.6 Sol
GPT-5.6 Sol has the lower estimated token cost for this stated workload. Claude Opus 4.8 has no published cached-input rate, so cached tokens use its listed input rate.
50K fresh input + 3K output tokens
GPT-5.6 Sol
GPT-5.6 Sol has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.
The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.
Like-for-like · BenchAlign v5.8
GPT-5.6 Sol has the higher coding point estimate. Conditional score ranges do not establish rank confidence.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
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.
OSWorld 2.0Agentic
Normalized gap 42.0FrontierMath v2 (Tier 4)Math
Normalized gap 26.9ARC-AGI-2Reasoning
Normalized gap 20.4FrontierMath v2 (Tiers 1-3)Math
Normalized gap 9.0BrowseCompAgentic
Normalized gap 7.9Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, reasoning, multimodal, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.
| Category | Claude Opus 4.8 | GPT-5.6 Sol | Basis | Reading |
|---|---|---|---|---|
| Agentic | 59.5Supported · #29/132 | 67.9Supported · #14/132 | Like-for-likeBenchAlign v5.8 lane · 14 vs 12 public rows | GPT-5.6 Sol leads · intervals overlap |
| Coding | 60.3Supported · #19/158 | 67.2Supported · #12/158 | Like-for-likeBenchAlign v5.8 lane · 12 vs 14 public rows | GPT-5.6 Sol leads · intervals overlap |
| Reasoning | 64.9Supported · #22/157 | 73.7Supported · #12/157 | Like-for-likeBenchAlign v5.8 lane · 3 vs 5 public rows | GPT-5.6 Sol leads · intervals overlap |
| Knowledge | 69.4Supported · #17/179 | 75.4Supported · #12/179 | Like-for-likeBenchAlign v5.8 lane · 6 vs 8 public rows | GPT-5.6 Sol leads · intervals overlap |
| Multimodal | 67.4Estimated · #18/67 | 75.4Supported · #9/67 | Directional onlyBenchAlign v5.8 lane · 6 vs 4 public rows | Directional only |
| Instruction following | 75.4#70/157 | 89.1#34/157 | Directional onlyProvisional lane · 0 vs 0 weighted rows | Directional only |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | 80.9#1/12 | 92.3Unranked · 3 rankable rows | Not comparableProvisional lane · 3 vs 2 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, reasoning, multimodal, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
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
GPT-5.6 Sol has the lower modeled cost
Costs use the listed standard API rates.
50K fresh input + 3K output tokens
GPT-5.6 Sol has the lower modeled cost
Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
GPT-5.6 Sol has the lower modeled cost
Claude Opus 4.8 has no published cached-input rate, so cached tokens use its listed input 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.
Claude Opus 4.8
GPT-5.6 Sol
1.05M
OpenAI model catalogClaude Opus 4.8
claude-opus-4-8
Anthropic model overviewGPT-5.6 Sol
gpt-5.6-sol
OpenAI model catalogA missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Claude Opus 4.8
Not published
GPT-5.6 Sol
$0.4 per 1M cached input tokens
OpenAI pricingClaude Opus 4.8
text, image
Anthropic model overviewGPT-5.6 Sol
text, image
OpenAI model catalogClaude Opus 4.8
GPT-5.6 Sol
Claude Opus 4.8
Generally Available · Claude API
Anthropic model overviewGPT-5.6 Sol
Generally Available · OpenAI Responses API
OpenAI model catalogClaude Opus 4.8
Reasoning
GPT-5.6 Sol
Reasoning
Claude Opus 4.8
Proprietary
GPT-5.6 Sol
Proprietary
Claude Opus 4.8
Proprietary
GPT-5.6 Sol
Proprietary
Claude Opus 4.8
2026-05-28
GPT-5.6 Sol
2026-07-09
GPT-5.6 Sol has the higher public score estimate, 74.36 versus 69.18, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
GPT-5.6 Sol has the higher public coding point estimate, 67.2 to 60.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
GPT-5.6 Sol has the higher public agentic tasks point estimate, 67.9 to 59.5, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
For the stated presets, chat costs $0.0175 on Claude Opus 4.8 and $0.014 on GPT-5.6 Sol; repository review costs $0.325 and $0.26; the cache-heavy agent loop costs $1.35 and $0.36. Claude Opus 4.8 has no published cached-input rate, so cached tokens use its listed input rate.
GPT-5.6 Sol has the larger documented context window: 1.05M, compared with 1M.
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Terminal-Bench 4.1
Shared sourceGPT-5.6 Sol leads this result
Terminal-Bench 3.0
Shared sourceGPT-5.6 Sol leads this result
Terminal-Bench 2.1
GPT-5.6 Sol leads this result
BrowseComp
GPT-5.6 Sol leads this result
DeepSearchQA
Not directly comparable
OSWorld-Verified
Not directly comparable
Finance Agent v2
Not directly comparable
MCP Atlas
Not directly comparable
Toolathlon
Claude Opus 4.8 leads this result
Gert Labs
Not directly comparable
ResearchClawBench
Not directly comparable
OSWorld 2.0
GPT-5.6 Sol leads this result
Terminal-Bench 2.1 (Vals)
GPT-5.6 Sol leads this result
MazeBench
GPT-5.6 Sol leads this result
LifeVal
Not directly comparable
CyberGym
Not directly comparable
ExploitGym
Not directly comparable
ApprenticeBench
Not directly comparable
SWE-bench Verified
Not directly comparable
SWE-bench Pro
Claude Opus 4.8 leads this result
SWE Multilingual
Not directly comparable
SWE Multimodal
Not directly comparable
Terminal-Bench 2.1
GPT-5.6 Sol leads this result
cursorBench31
Not directly comparable
CursorBench 3.2
Shared sourceGPT-5.6 Sol leads this result
FrontierCode 1.1 Main
Not directly comparable
LiveCodeBench (Vals)
Claude Opus 4.8 leads this result
SWE-bench (Vals)
GPT-5.6 Sol leads this result
PostTrainBench v1.1
Shared sourceGPT-5.6 Sol leads this result
FrontierCode 2.0
Shared sourceGPT-5.6 Sol leads this result
Bug Hunt Bench
Not directly comparable
DeepSWE
Not directly comparable
FrontierCode 1.1 Extended
Not directly comparable
FrontierSWE v2
Not directly comparable
VulcanBench v3
Not directly comparable
VulcanBench CII v1
Not directly comparable
CursorBench 4.0
Not directly comparable
ARC-AGI-2
GPT-5.6 Sol leads this result
ARC-AGI-3
GPT-5.6 Sol leads this result
ARC-AGI-1
GPT-5.6 Sol leads this result
BioBasicsBench
Not directly comparable
GeneBench-Pro
Not directly comparable
ExtractBench (LlamaIndex)
Shared sourceGPT-5.6 Sol leads this result
ParseBench
Shared sourceGPT-5.6 Sol leads this result
OfficeQA Pro
Not directly comparable
ScreenSpot Pro
Not directly comparable
CharXiv
Not directly comparable
CharXiv w/o tools
Not directly comparable
MMMU-Pro
Not directly comparable
MMMU-Pro w/ Python
Not directly comparable
GPQA
GPT-5.6 Sol leads this result
GPQA-D
GPT-5.6 Sol leads this result
HLE
Not directly comparable
HLE w/o tools
Not directly comparable
GPQA Diamond (Vals)
GPT-5.6 Sol leads this result
MMLU-Pro (Vals)
Claude Opus 4.8 leads this result
HLE-Verified
Not directly comparable
LABBench2
Not directly comparable
HealthBench Professional
Not directly comparable
HealthBench Hard
Not directly comparable
INCLUDE
Not directly comparable
Gray Swan IPI (15 attempts)
Not directly comparable
USAMO 2026
Not directly comparable
FrontierMath v2 (Tiers 1-3)
GPT-5.6 Sol leads this result
FrontierMath v2 (Tier 4)
GPT-5.6 Sol leads this result
FrontierMath (legacy)
Not directly comparable
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 6,100+ subscribers.
Last updated November 9, 2026
