Kimi-K3 Predicted to Score 70.3 on ARC-AGI-2 Benchmark

BenchPress predicts the Kimi-K3 model will achieve a score of 70.3 on the ARC-AGI-2 benchmark. This projected performance places Kimi-K3 between Anthropic's Opus 4.6 and OpenAI's GPT-5.4-xhigh models.

So What

A new model's predicted performance on a rigorous benchmark provides an early signal of its competitive standing against frontier models.

View Tweet→

Sources (1)

Kimi-K3 Predicted to Score 70.3 on ARC-AGI-2 Benchmark

General AI Twitter · Twitter · 4d ago

View tweet→

Related Signals

5d agoModel ReleaseKimi.ai Introduces Kimi K3 Open Frontier Intelligence Model→3d agoBenchmarkKimi-K3 Achieves Preliminary ECI Score of 155.53, Surpassing Opus 4.6 and GPT-5.3-Codex→6h agoBenchmarkEpoch AI Reports Kimi K3 Sets Open-Weights Record on Capabilities Index→5d agoBenchmarkKimi-K3 Scores 57 on AAI Index, Demonstrates 2x Token Efficiency Over Opus 4.8→4d agoBenchmarkKimi K3 Ranks #5 on Artificial Analysis Coding Agent Index with Cost-Efficient Performance→

Credibility

Lead sourceGeneral AI Twitter

Corroboration1 distinct outlet

VerificationSingle source

Entities

Anthropicrelated OpenAIrelated

Unlock the full signal with Pro

Tessara synthesizes 80+ data sources into actionable intelligence for AI infrastructure investors.

Unlock with Pro→Sign in

← SignalsSignal

Watch7.5BenchmarkModelsSingle sourceactive

Updated 4d ago · 1 source

Kimi-K3 Predicted to Score 70.3 on ARC-AGI-2 Benchmark

BenchPress predicts the Kimi-K3 model will achieve a score of 70.3 on the ARC-AGI-2 benchmark. This projected performance places Kimi-K3 between Anthropic's Opus 4.6 and OpenAI's GPT-5.4-xhigh models.

So What

A new model's predicted performance on a rigorous benchmark provides an early signal of its competitive standing against frontier models.

View Tweet→