LLM Guides Lab

LLM Leaderboard: ChatGPT vs Claude vs Gemini

ChatGPT vs Claude vs Gemini, ranked on real everyday tasks.

The first quality tests are in. Rankings start once a model has results in three quality tests; until then every model is listed as provisional.

Updated October 8, 2026 1 quality test 24 answers, all public Scored blind

Provisional standingsLab Index · out of 5

–

ClaudeLeads 1 of 1 test

4.8
–

ChatGPTValue

4.6
–

Gemini1 quality test so far

4.1

"=" shared position: gap under 0.2Full ranking ↓

Full ranking

The LLM Guides Lab Index is the average of each model’s latest score in every quality test.

LLM Guides Lab ranking of models, October 2026
#ModelLab IndexTrendChangeTestsBest atCost / point
–
Claude ProvisionalOpus 5.5
 4.8 New 1 / 1 Writing $0.0122
–
ChatGPT ProvisionalGPT-6 Sol
 4.6 New 1 / 1 Value $0.0033
–
Gemini Provisional3.8 Flash
 4.1 New 1 / 1 $0.0012

Scores out of 5 · a lead under 0.2 counts as a tie · ranked after 3 quality tests · cost per point = API cost of the test answers ÷ scoreHow we score →

Today’s winner isn’t guaranteed tomorrow.

The models update every few weeks, and we retest when they do. Get the new rankings the day they shift.

No spam. Unsubscribe anytime. Privacy policy

Trends and value

LLM Guides Lab Index over time

Average across quality tests, out of 5

ClaudeChatGPTGemini

5.04.64.23.8 Oct 2026 4.8 4.6 4.1
Lab Index by date
ModelOctober 8, 2026 (not directly comparable)
Claude4.8
ChatGPT4.6
Gemini4.1

Dashed lines: fewer than three quality tests existed at that date, so those points aren’t directly comparable. Each test’s full history is on its own page.

Best value

API cost per quality point. Lower is better.

Gemini$0.0012
ChatGPT$0.0033
Claude$0.0122

ChatGPT reaches 96% of Claude’s Index for about 26% of the cost.

Best model for each task

The latest score in every test. Outlined = winner; "=" marks a shared top spot (lead under 0.2).

Latest score per model in each LLM Guides Lab test
TestClaudeChatGPTGeminiLast tested
Writing8 prompts · tested once 4.8 4.6 4.1 Oct 2026

How the leaderboard works

Four rules, applied the same way every time.

Equal weight per test

Each quality test counts the same, however many prompts it has.

Latest results only

The Index uses each test’s most recent date. Older results stay on the test pages.

Ties are ties

A lead under 0.2 is shown as a shared position, because single runs vary.

3 tests to be ranked

New models are listed as provisional until they’ve been through three quality tests.

Every score comes from blind, hands-on testing with published prompts and answers. Read the full methodology →

Use the data

Download this leaderboard as CSV, or browse every answer and score on GitHub.

Download CSV All data on GitHub

Citing the leaderboard: LLM Guides Lab, "LLM Leaderboard," llmguides.ai/lab/leaderboard/, updated October 2026.

Data licensed CC BY 4.0: free to reuse with credit to LLM Guides.