Which model leads Terminal-bench 3.0?

Q: Which model leads Terminal-bench 3.0?

GPT-5.6 Sol tops the board at 34.4

Terminal-bench 3.0 Leaderboard — 14 Models

Pro Mode — raw scores, one benchmark under the lensofficial card harness

The field · 12 measured

4.634.4

12 models measured, most on the official card harness. GPT-5.6 Sol tops the board at 34.4.

12 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 12Bars show 4.6–34.4 · ◆ solid = first-party or better · outlined = arm's-length

01GPT-5.6 Sol+2 altsOpenAI34.4$4.0/M$20.0/M$0.349

02DeepSeek-V4.1-FlashDeepSeek30$0.20/M$0.60/M$0.013

03GLM-5.3+3 altsZ.ai28.3$1.4/M$4.4/M$0.102

04Grok 4.6SpaceXAI26$2.0/M$6.0/M$0.154

05Claude Opus 4.8+1 altAnthropic21.13P$5.0/M$25.0/M$0.711

06GPT-5.6 TerraOpenAI20.8$2.0/M$12.0/M$0.337

07Kimi K3+1 altMoonshot17.43P$0.58/M$12.3/M$0.370

08Gemini 3.7 Flash+2 altsGoogle14.9$0.75/M$3.8/M$0.151

09Claude Sonnet 5Anthropic14.6$2.0/M$10.0/M$0.411

10DeepSeek-V4-ProDeepSeek11.8$0.43/M$0.87/M$0.055

11Gemini 3.6 FlashGoogle5.4$0.75/M$3.8/M$0.417

12GLM-5.2+1 altZ.ai4.63P$1.4/M$4.4/M$0.630

◆ solid = independently verified · aggregator attested · vendor attributed — outlined = cross-reference · source-attributed. Dimmed ◆ = confidence inferred. “+N alts” = alternate disclosures in the drawer with why-demoted. Prices per million tokens; $/point = input price ÷ score. Every row opens its model dossier.

Verification: 12 models scored · 0 independently verified · 5 vendor cross-reference · 7 vendor-reported · 0 with source disagreement. How these tiers are assigned