ReasoningActive
Trust Score:
93

CritPt v2 (Hard)

Hard subset of version 2 of the CritPt research-level physics benchmark.

Launched: —Refresh: static
Status Assessment (active):

Actively evaluated by Artificial Analysis on current frontier models; no saturation evidence recorded.

Performance Timeline

Longitudinal progression of model scores against human baselines.

Performance & Historical Trajectory

Empirical score progression across model release dates and evaluation rounds.

Independent Vendor-Reported
ModelScoreDateSource TypeProvenance
NVIDIA Nemotron 3 Nano 4B0%2026-11-15independentSource ↗
Gemini 4 Argon52.5%2026-11-15independentSource ↗
Nanbeige4.3-33B15.5%2026-11-15independentSource ↗
Ling 3.1 Flash 101938.5%2026-11-15independentSource ↗
K2 Horizon MoVA 36B A4B8%2026-11-15independentSource ↗
DeepSeek V4.1 Flash 101725%2026-11-15independentSource ↗
Muse Glimmer7%2026-11-15independentSource ↗
Nemotron 3 Ultra15%2026-11-15independentSource ↗
Kimi K3.1 Pro49.5%2026-11-15independentSource ↗
Beam 1.026.5%2026-11-15independentSource ↗
Inkling Small18.5%2026-11-15independentSource ↗
Gemini 3.8 Flash46.5%2026-11-15independentSource ↗
Muse Ember39%2026-11-15independentSource ↗
DeepSeek V4-Pro23.5%2026-11-15independentSource ↗
Gemini 4 Hydrogen66%2026-11-15independentSource ↗
Qwen3.8-Flash-Next24.5%2026-11-15independentSource ↗
Muse Wonder50.5%2026-11-15independentSource ↗
Granite 4.2 8B0%2026-11-15independentSource ↗
Qwen3.8 Max (0902)29.5%2026-11-15independentSource ↗
JT-4.1 Flash 236B A21B19%2026-11-15independentSource ↗
K2 Horizon 375B A23B12%2026-11-15independentSource ↗
gpt-oss-120b4%2026-11-15independentSource ↗
GLM 5.333.5%2026-11-15independentSource ↗
Gemma 4 31B4%2026-11-15independentSource ↗
Hy439%2026-11-15independentSource ↗
MiniMax M3.1 Flash32.5%2026-11-15independentSource ↗
Agnes 2.5 Pro Theta26%2026-11-15independentSource ↗
Mistral Medium 3.50.5%2026-11-15independentSource ↗
Nemotron 3.5 Super VL13.5%2026-11-15independentSource ↗
Inkling 2 Small35%2026-11-15independentSource ↗
Muse Wonder Full56.5%2026-11-15independentSource ↗
GLM 5.3 Flash24.5%2026-11-15independentSource ↗
Apodex 1.117%2026-11-15independentSource ↗
GPT-6.1 Sol69.5%2026-11-15independentSource ↗
Gemini 3.5 Flash-Lite3.5%2026-11-15independentSource ↗
Step 532.5%2026-11-15independentSource ↗
Granite 4.2 3B0%2026-11-15independentSource ↗
Claude Fable 5.276.5%2026-11-15independentSource ↗
GPT-6.1 Terra68%2026-11-15independentSource ↗
Gemini 4 Carbon69%2026-11-15independentSource ↗
Inkling18%2026-11-15independentSource ↗
Granite 4.2 30B1.5%2026-11-15independentSource ↗
K2 Horizon 7B1.5%2026-11-15independentSource ↗
MiMo-V2.6-Flash29.5%2026-11-15independentSource ↗
Mistral Small 40%2026-11-15independentSource ↗
Solar Pro 418%2026-11-15independentSource ↗
K-EXAONE 2.04%2026-11-15independentSource ↗
Claude Sonnet 5.559%2026-11-15independentSource ↗
GPT-6 Luna42.5%2026-11-15independentSource ↗
North Mini Code1%2026-11-15independentSource ↗
GPT-6 Astra70.5%2026-11-15independentSource ↗
Claude Haiku 5.543.5%2026-11-15independentSource ↗
Grok 4.733.5%2026-11-15independentSource ↗
MiMo-V2.6-Pro34%2026-11-15independentSource ↗
MiniMax M328%2026-11-15independentSource ↗
Mistral Large 427%2026-11-15independentSource ↗
Nemotron 3.5 Lightning1.5%2026-11-15independentSource ↗
Command A+2.5%2026-11-15independentSource ↗
Claude Opus 5.568%2026-11-15independentSource ↗
K2 Horizon 3.7B1%2026-11-15independentSource ↗
Qwen3.8 27B4%2026-11-15independentSource ↗
Muse Spark Open29.5%2026-11-15independentSource ↗

Human Baseline & Difficulty Horizon

Calibrated human reference points, specialist benchmarks, and ceiling thresholds.
Measured Human Scoren/aNot established
Baseline Protocol & Interpretation

No peer-reviewed human baseline has been published for this benchmark.

Metric & Scoring Methodology

Verification protocols, aggregation formulas, and specialized metric variants.
Primary Metric:accuracy (%)
Scoring Engine:exact-match

Dataset & Compute Cost

Evaluation volume, public availability, API pricing, and local hardware requirements.
Total Dataset Size—Annotated evaluation items
Public Test SetPublicOpenly mirrored on repositories
Access GatingOpen AccessUnrestricted download
Evaluation LicenseOpen AccessDataset usage and redistribution terms
Frontier API Compute Cost:

$5 – $20 USD for full benchmark evaluation run on frontier APIs.

Recommended Local GPU Setup:

1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang

How to Run & Reproduce

Standardized evaluation protocols, CLI commands, and reproducible runner templates.
Prompt Regimezero-shot
Reasoning Modedirect
Sampling Temp0
Pass@k Budgetk = 1
Tools & SandboxPure Text
Scoring Verifierexact-match
Option AEleutherAI LM-Evaluation-Harness (Open-Weight Models)
lm_eval --model hf --model_args pretrained=<model_path> --tasks critpt-v2 --batch_size auto
Option BOpenCompass Evaluation Framework
opencompass --datasets critpt-v2 --models <model_config>
Python APIDeterministic Inference Loop Snippet
# Standard API Evaluation Loop
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}],
    temperature=0.0,
)
Standardized Reporting Requirement:

When publishing results for CritPt v2 (Hard), always report the exact prompt template, few-shot exemplar ordering, sampling temperature (temperature=0), maximum reasoning budget tokens, and the precise timestamped model snapshot ID.

Contamination & Memorization Analysis

Audit of pretraining exposure risks, memorization vectors, and refresh policies.
Overall Contamination Risk:MEDIUM
Refresh Cadence:

Static fixed snapshot

Test Set Exposure:

Public on web / HuggingFace