How many entities can I compare at once?

Two to four of the same type. Mixing models with agents or providers is not supported · each comparison works on a single entity class.

What do BenchGecko comparisons include?

For models · benchmark bars, full benchmark table, pricing per 1M tokens, projected monthly cost, Hype vs Reality, Value, and Vendor Risk callouts. Agents and providers add GitHub stars velocity, pricing, speed, and mindshare where available.

Can I share a comparison URL?

Yes. Every comparison has a permanent URL like /compare/claude-opus-4-6-vs-gpt-5-chat. Copy the URL from the address bar to share. Comparisons render on-demand and re-render with the freshest data we have.

The Matrix is the full scoring grid · every model against every benchmark in a single sortable table. Open /compare/matrix for the grid view.

Why do some comparisons show dashes?

A dash means we do not have that data point for that entity. Benchmark coverage is uneven · we only show rows where at least two entities have a score.

Compare · UniversalLive · 971 models · 25 agents · 268 providers · 71 companies

Compare Everything · Side by Side

Benchmarks, pricing, mindshare, attention premium. Build your own comparison or pick from our curated sets.

Cite this page Share

Instant answers

12 questions · 12 pre-built comparisons · one tap

Claude vs GPT-5?

winner

Claudeon 7 of 10 benchmarks

Best for coding?

winner

Claude Codeon SWE-bench + dev choice

Cheapest capable?

winner

DeepSeek V3.2on cost per token

Best OSS?

winner

Qwen3.5on top OSS composite

Agent wars?

winner

Claude Codeon GitHub stars velocity

Frontier models?

winner

Claude Mythoson reasoning + coding

Big Tech race?

winner

Googleon valuation · capex

Under $1 per M?

winner

DeepSeek V3.2on cheapest tier

Who makes chips?

winner

NVIDIAon market share · AICDI

DeepSeek hype?

winner

DeepSeekon Hype vs Reality

Open source top 5?

winner

Qwen3.5on composite score

Vision champs?

winner

Gemini 3.1on Video-MME · multimodal

Starter packs

curated by use case · 4 entities each

The four models that actually ship production code.

tested on SWE-bench + HumanEval + Terminal-bench

Who actually thinks · GPQA, AIME, MATH.

tested on GPQA + AIME + MATH-500

GitHub stars · harness type · developer adoption.

tracked on stars velocity + dev choice index

Best score per dollar · run anywhere.

scored per 1M input tokens

Battles of the week

trending comparisons · last 7 days

The Matrix · full grid

Every model · every benchmark

The full scoring grid in a single sortable view. Filter by provider, pricing, or context. Screenshot and share.

Open the matrix

Or · browse the indexes

check 2–4 on any page · pink Compare button · done

How it works

3 steps · no login · no setup

Browse or build

Use the Compare Builder above or open any index · /models, /agents, /providers, /economy/companies.

Select 2 to 4

Check 2 to 4 entities of the same type. A floating compare bar appears at the bottom.

Compare

Benchmark bars, pricing, mindshare, vendor risk side by side. Permanent shareable URL.

Compare FAQ

Use the Compare Builder above. Pick an entity type · Models, Agents, Providers, or Companies · check 2 to 4 items from the list, then hit Compare. For models you can also focus the comparison by category or specific benchmark.

Compare Everything · Side by Side

Custom Compare Builder

Pick what you want to compare

Pick 2 to 4 to compare

Focus the comparison

Instant answers

Starter packs

Battles of the week

Or · browse the indexes

How it works

Compare FAQ