Filled points are supported; hollow points are preliminary. Lines stop at unknown capabilities. Exact values and sources follow below.
Agentic6 families · 0 with independent evidence · Supported34.0
6 core families; 11 direct opponents across 3 labs.
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: 28.4–39.9; 0/13 scenarios unsupported. Without one publisher: 18.8–43.9; 0/15 unsupported. Smoothing check: 31.1–38.7; 0/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage; source sensitivity still requires inspection. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
- GDPval: 38.6 observed win share
Google, OpenAI · Source 1 Source 2 - GDP.pdf: 35.0 observed win share
Google, OpenAI · Source 1 Source 2 - Browsecomp: 80.0 observed win share
OpenAI · Source 1 - AutomationBench: 50.0 observed win share
OpenAI · Source 1 - OSWorld: 50.0 observed win share
OpenAI · Source 1 - Toolathlon: 16.7 observed win share
OpenAI · Source 1
Hard reasoning4 families · 1 with independent evidence · Supported52.2
4 core families; 61 direct opponents across 10 labs.
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: 46.3–58.8; 0/6 scenarios unsupported. Without one publisher: 32.6–53.2; 0/16 unsupported. Smoothing check: 51.3–53.0; 0/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
Coding3 families · 2 with independent evidence · Supported56.3
3 core families; 35 direct opponents across 9 labs.
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: 48.7–66.6; 0/6 scenarios unsupported. Without one publisher: 54.3–60.1; 0/19 unsupported. Smoothing check: 55.2–56.3; 0/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
Human prefNo comparable evidenceUnknown
0 core families; 0 direct opponents across 0 labs. No admitted core comparisons
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: No supported estimate; 1/1 scenarios unsupported. Without one publisher: No supported estimate; 1/1 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: cross-family and cross-publisher predictions could not be evaluated from the available comparison network. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
KnowledgeNo comparable evidenceUnknown
0 core families; 0 direct opponents across 0 labs. No admitted core comparisons
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: No supported estimate; 4/4 scenarios unsupported. Without one publisher: No supported estimate; 5/5 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: almost no publisher-held-out comparisons remained estimable. Cross-source validity is unresolved. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
Multimodal3 families · 0 with independent evidence · Supported48.9
3 core families; 8 direct opponents across 3 labs.
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: 46.9–61.5; 0/8 scenarios unsupported. Without one publisher: 35.1–66.8; 1/8 unsupported. Smoothing check: 48.0–49.7; 0/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: limited publisher-held-out coverage. Cross-source validity is unresolved. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
Long context2 families · 0 with independent evidence · Supported50.9
2 core families; 4 direct opponents across 2 labs.
Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.
Refit without one family: 50.9–51.5; 2/5 scenarios unsupported. Without one publisher: 50.9–51.7; 2/5 unsupported. Smoothing check: 48.4–53.9; 0/3 unsupported. These are sensitivity checks, not confidence intervals.
Audit: family and publisher holdouts underperformed neutral predictions. Treat this as an exploratory estimate. Validation audit.
Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.
Scores estimate outcomes against a shared reference panel; they are not accuracy percentages. Sparse or disconnected evidence cannot qualify an overall profile. Open a capability to inspect its evidence.
/badge/gpt-5.6-terra.svg