GPT-5.6 Sol vs GPT-6 Sol
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
GPT-5.6 Sol51.3
GPT-6 Sol58
GDPval-AA (work tasks) 2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work
GPT-5.6 Sol46.1
GPT-6 Sol42.3
ITBench SRE ITBench SRE: IT incident tasks: diagnose and fix broken systems like a site-reliability engineer
GPT-5.6 Sol · max56.2
GPT-6 Sol · max49.4
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
GPT-5.6 Sol14.6
GPT-6 Sol18.7
Terminal-Bench Science 0.1 Terminal-Bench Science 0.1: Agents complete expert-authored research workflows across five scientific domains in a terminal
GPT-5.6 Sol · max22.4
GPT-6 Sol8.6
AA-Briefcase 1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating
GPT-5.6 Sol1,241.9
GPT-6 Sol1,149.3
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
GPT-5.6 Sol80.3
GPT-6 Sol82.3
DeepSWE 1.1 DeepSWE 1.1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition
GPT-5.6 Sol61.1
GPT-6 Sol · max70.5
EBR Bench EBR Bench: Agents learn to play Earthborne Rangers across repeated playthroughs
GPT-5.6 Sol · max44.8
GPT-6 Sol · max53.3
Furniture Assembly Furniture Assembly: Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step
GPT-5.6 Sol · max56.7
GPT-6 Sol · max58.3
τ³-bench · banking · 1.0.1 τ-bench official · 3-banking-1.0.1: Customer-service agents converse with simulated users, retrieve domain knowledge and change state through tools
GPT-5.6 Sol · xhigh46.9
GPT-6 Sol—
τ³-Banking · AA τ³-Banking · AA: Agents retrieve banking knowledge and execute tools to resolve customer-support workflows
GPT-5.6 Sol36.5
GPT-6 Sol—
AA Analyst Agent AA Analyst Agent: Data-analyst tasks: the model explores data with tools and reports findings
GPT-5.6 Sol · max47.5
GPT-6 Sol—
Terminal-Bench Hard (version unstated) Terminal-Bench Hard (version unstated): Agents solve difficult software, system-administration and data-processing tasks in a terminal
GPT-5.6 Sol62.9
GPT-6 Sol—
EnterpriseOps-Gym-AA EnterpriseOps-Gym-AA: Agents complete stateful enterprise workflows in eight domains; final database state determines success
GPT-5.6 Sol · max42.9
GPT-6 Sol—
Harvey LAB-AA 1.1 Harvey LAB-AA 1.1: Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit
GPT-5.6 Sol—
GPT-6 Sol · max3.6
SWE-bench Pro (v2 full) SWE-bench Pro (v2 full): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate
GPT-5.6 Sol · xhigh95.5
GPT-6 Sol—
SWE-bench Pro (v2 hard) SWE-bench Pro (v2 hard): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate
GPT-5.6 Sol · xhigh82.4
GPT-6 Sol—
SpireClimb SpireClimb: Agents play Slay the Spire under the named game and difficulty setting
GPT-5.6 Sol—
GPT-6 Sol · max7.5
SpireClimb StS1 A20 Heart SpireClimb StS1 A20 Heart: Agents play Slay the Spire under the named game and difficulty setting
GPT-5.6 Sol—
GPT-6 Sol · max1.7
SpireClimb StS2 A10 SpireClimb StS2 A10: Agents play Slay the Spire under the named game and difficulty setting
GPT-5.6 Sol—
GPT-6 Sol · max13.3
| Benchmark | GPT-5.6 Sol | GPT-6 Sol |
|---|---|---|
| AutomationBench-AA | 51.3 | 58 |
| GDPval-AA (work tasks) 2.1 | 46.1 | 42.3 |
| ITBench SRE | 56.2 · max | 49.4 · max |
| Terminal-Bench 4.0 | 14.6 | 18.7 |
| Terminal-Bench Science 0.1 | 22.4 · max | 8.6 |
| AA-Briefcase 1.1 · Elo | 1,241.9 | 1,149.3 |
| AA-LCR (long context) 1.1 | 80.3 | 82.3 |
| DeepSWE 1.1 | 61.1 | 70.5 · max |
| EBR Bench | 44.8 · max | 53.3 · max |
| Furniture Assembly | 56.7 · max | 58.3 · max |
| τ³-bench · banking · 1.0.1 | 46.9 · xhigh | — |
| τ³-Banking · AA | 36.5 | — |
| AA Analyst Agent | 47.5 · max | — |
| Terminal-Bench Hard (version unstated) | 62.9 | — |
| EnterpriseOps-Gym-AA | 42.9 · max | — |
| Harvey LAB-AA 1.1 | — | 3.6 · max |
| SWE-bench Pro (v2 full) | 95.5 · xhigh | — |
| SWE-bench Pro (v2 hard) | 82.4 · xhigh | — |
| SpireClimb | — | 7.5 · max |
| SpireClimb StS1 A20 Heart | — | 1.7 · max |
| SpireClimb StS2 A10 | — | 13.3 · max |
Specs
| Spec | GPT-5.6 Sol | GPT-6 Sol |
|---|---|---|
| Input, per 1M | $4 | $2 |
| Output, per 1M | $20 | $10 |
| Cost per task? | $0.50 | $0.25 |
| Speed, tok/s | 85 | — |
| Context | 1.05M | 1.05M |
| Weights | Closed | Closed |
| Released | Jul 2026 | Sep 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | Partial |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
GPT-5.6 Sol → GPT-6 Sol
Nothing known.
GPT-6 Sol → GPT-5.6 Sol
Nothing known.