League

The same real-life cases for every model, scored against rules fixed in advance. Every answer and verdict is open to inspect.

Season 1 · 10 cases · 6 models · publication details

Battery:: rl-1 (frozen revision 4)
Published:: 18 Oct 2026
Corrected:: 19 Oct 2026
Evaluation:: reallife-check-2

Season 1

Standings

10/30rounds published

1=DeepSeek V4 Pro0 critical · 10 cases39/3939 pts+3 this round
1=Gemini 4 Argon0 critical · 10 cases39/3939 pts+3 this round
1=GPT-6 Astra0 critical · 10 cases39/3939 pts+3 this round
1=Grok 4.70 critical · 10 cases39/3939 pts+3 this round
5=Claude Sonnet 50 critical · 10 cases38/3938 pts+3 this round
6=Mistral Large3 critical · 10 cases34/3928 pts−1 this round

#	Model	Checks passed	Critical misses	Cases	Points
1=	DeepSeek V4 ProDeepSeek	39/39	0	10	39+3 this round
1=	Gemini 4 ArgonGoogle	39/39	0	10	39+3 this round
1=	GPT-6 AstraOpenAI	39/39	0	10	39+3 this round
1=	Grok 4.7xAI	39/39	0	10	39+3 this round
5=	Claude Sonnet 5Anthropic	38/39	0	10	38+3 this round
6=	Mistral LargeMistral	34/39	3	10	28−1 this round

4 models share full marks (39/39 checks). Claude Sonnet 5 is one check behind. Positions describe only these 10 shared tasks after 10 rounds, not overall model quality.

Rules, fixed in advance

01Only scored challenges count. Perspectives never earn points.
02Models are compared only on tasks they all ran.
03Points = checks passed − 2 per critical failure.
04Equal points share a position. Speed and cost never affect points.
05The table updates after each published round: “+3 this round” is the latest round alone (checks passed − 2 per critical miss); arrows show places moved.

All product names, logos and brands are property of their respective owners. They are used on this site for identification and comparative benchmarking only; their use does not imply affiliation with or endorsement by them.

League · Season 1 · Published 18 Oct 2026

Season 1 evidence: every case, answer and verdict

10 cases · 6 models. Open a case to read the task, then any model's answer and check results.

Fingerprint

Battery rl-1-draft-4 · SHA-256 b954319990d54d90e4b17499e1e3d26ba085265d586068f0367f69e49d381f5a

Only failed checks

01Tenant repair request6/6 full pass
02Monthly household budget5/6 full pass
03Doctor's appointment summary6/6 full pass
04Holiday handover note6/6 full pass
05Refund complaint for a faulty product6/6 full pass
06Cloud migration quote comparison4/6 full pass
07API deprecation notice6/6 full pass
08Security incident status update6/6 full pass
09Three-city meeting window6/6 full pass
10Travel expense policy audit5/6 full pass

League guide

How to read the League

A short email guide: how rounds are scored, what the points and arrows mean, and how to check any result yourself.

Season standings

Standings

How to read the League