AI benchmarks
10 articles tagged “AI benchmarks”.
Tagged content
Search everything-
Read →Gemini 4 Argon scores 97.5% on ARC-AGI-3, 49.8% on standard harness
ARC Prize's verified ARC-AGI-3 results for Gemini 4 Argon: 97.5% with the Provider Adapter harness, 49.8% with the standard harness, and what the Dec…
-
Read →Muse Wonder Is Out: Where to Use It and How It Benchmarks
Meta's Muse Wonder is available in the Meta Model API and Muse Code, and Muse Wonder Full is for Muse Code subscribers. Here is who can use each,…
-
Read →Xiaomi MiMo-V2.6 Agent Arena Rankings: Pro vs Flash
Arena reports different strengths for Xiaomi’s MiMo-V2.6-Pro and MiMo-V2.6-Flash, with Pro ranking higher and Flash costing less per task.
-
Read →Gemini 4 Argon: official announcement, then leaks and reports on access and price
Google has announced Gemini 4 Argon for trusted testers first, with no general access date. Earlier Gemini 4 reports remain unverified.
-
Read →What OpenAI Says MentalHealthBench Evaluates
OpenAI describes MentalHealthBench as an openly released benchmark for mental-health conversations, from everyday support to acute crises. Its…
-
Read →Claude Opus 5.5 Early Review: First Impressions on Speed, Pricing and Coding
Claude Opus 5.5 review: first impressions on coding, 3D game development, speed, API pricing, benchmarks and usage limits.
-
Read →What Changed in SWE-Bench Pro V2?
Scale AI Labs says SWE-Bench Pro V2 refreshes its public task set and adds checks on task validity, agent access and solution grading—while leaving…
-
Read →Xiaomi MiMo-V2.6 Pro and Flash: Reported Benchmarks
Xiaomi announces MiMo-V2.6 Pro and Flash with open model weights and reports mixed benchmark results against MiMo-V2.5 and named frontier models.
-
Read →Grok 4.7 timeline: from release countdown to xAI's official announcement
xAI has announced Grok 4.7 and listed its prices. Earlier leaks and sightings are tracked here, with company-reported benchmark results kept separate…
-
Read →Cognition’s SWE-2 Targets Better Coding Performance at Lower Cost
Cognition’s SWE-2 adds medium, high, and max effort levels, with company-reported gains in coding benchmarks and efficiency over SWE-1.7.