UABUnbiased AI BenchAI model rankings with source links.
Every score links back to its source.
Home/Compare
Compare
Live · updated continuously
Compare · pick a model

Compare the models people actually argue about.

Use shared tests, visible gaps, and close-call warnings before turning a benchmark lead into a public claim.
Selected models · 1
Shared benchmarks · 5
Current leader · kimi-k2.5-thinking
Compare controls

Keep the claim honest.

Use case and source filter stay visible up top. Edits, saves, follows, and links stay available without running the page.

Current frameBest for this use caseCoding copilot · All public sources
What should win?
Preset

Keep the same models, change the use case, and recompute the recommendation without resetting the page.

Sources
Models to compare

Keep the compare set tight. Add one model at a time, then prune aggressively.

1 / 6 models
Workspace utilities

Save, follow, or copy this exact setup only after the comparison looks right.

kimi-k2.5-thinking
Save or share this workspacefollow it, copy it, or reopen it later
Open evidence pageOpen workspaces
Recommendation

kimi-k2.5-thinking leads this compare set for coding copilot.

Read this page as a workspace for choosing between top models, not as a universal crown. The current frame is Coding copilot under best for this use case.

Thin verified coverageThe source data is visible, but a direct ranking is blocked by these rules: benchmark_derived_model
Leader for this use casekimi-k2.5-thinking
Closest objectionNo clear runner-up yet
Shared tests5 shared benchmarks · 0 close calls
SourcesAll public sources
Why the leader is ahead
Chat / text · Vision understanding
Where the evidence is thin
0 close-call benchmarks and 35 missing missing leader rows keep this from reading as settled.
What to do next
Inspect decisive benchmarks first, then open the disagreement page or head-to-head pages if the top line still feels too narrow.
Open evidence pageInspect benchmark disagreement
5 of 40 benchmarks
Text Arena
AR · rating
Text · Chat / text
1,445n/a
n/a
Code Arena
AR · rating
Code · Coding
1,431n/a
n/a
Vision Arena
AR · rating
Vision · Vision understanding
1,248n/a
n/a
WebDev Arena
AR · rating
Code · Coding
1,431n/a
n/a
Document Arena
AR · rating
Document · Document understanding
1,437n/a
n/a
Decisive benchmarks

What is doing the visible work

No decisive benchmark
The shared tests are too split or too sparse for a clean separator.
Decision pressure

What changes the winner

kimi-k2.5-thinking
35 visible benchmark gaps still leave room for the result to move.
Coverage gaps

Where evidence is missing

kimi-k2.5-thinking
Missing visible evidence on this compare surface
35
Preset interpretation

How this weighting reads the field

The current decision mode is grounded in the Coding copilot preset. This keeps the compare page connected to a visible use case instead of an unspoken “overall winner” claim.

score = 1.45 × coding + 1.05 × reasoning math science + 0.80 × long contextcoverage floor = 55% · recency window = 120 days
Share or publish this compare resultlinks, copy, and advanced framings
Share this evidence page

Post the claim with the evidence attached.

Use the evidence page for the full source trail, or the card image when the post needs a clean preview.

Model comparekimi-k2.5-thinking leads this compare set for coding copilot.

0 shared benchmarks are still too close to call, so the win stays conditional. This compare uses all public sources, with provider-official evidence labeled separately.

Evidence page/artifacts/compare?models=kimi-k2-5-thinking&mode=best-for-this-use-case&preset=coding-copilot
Card image/api/og?kind=compare&models=kimi-k2-5-thinking&mode=best-for-this-use-case&preset=coding-copilot
Public links

Open or copy the stable surfaces

The evidence page is the canonical source surface. The card image is the compact preview for embeds, screenshots, and social cards.

Open evidence pageOpen card preview
Copy-ready text

Use the exact public framing

Each copy action keeps the claim attached to evidence instead of forcing you into a blank composer.

Advanced framings and X composerNeutral, contrarian, open-model, and skeptical variants
Model compare

Pick the voice before you post

Use the framing variants only when you need them. The evidence page and the public copy actions above should handle most cases.

Neutral analystLead with the claim, then attach the reason and warning.kimi-k2.5-thinking leads this compare set for coding copilot.
ContrarianPush against the easy read and keep the strongest alternative live.Contrarian take: kimi-k2.5-thinking leads this compare set for coding copilot.
Open-model angleBias the framing toward the open-weight or transparent-evidence angle.Open-model angle: Model compare · kimi-k2.5-thinking
Don't trust the headlineLead with the warning before you let the claim travel.Don't trust the headline: Model compare · kimi-k2.5-thinking
X composer

Compose a post that keeps the warning attached

The post shell always exposes the claim, why, warning, evidence link, and an optional discussion question.

Headlinekimi-k2.5-thinking leads this compare set for coding copilot.
WhyThe visible source data changed enough to change the claim.
Warning0 shared benchmarks are still too close to call, so the win stays conditional. This compare uses all public sources, with provider-official evidence labeled separately.
Evidence link/artifacts/compare?models=kimi-k2-5-thinking&mode=best-for-this-use-case&preset=coding-copilot
PreviewOver 280
kimi-k2.5-thinking leads this compare set for coding copilot.
The visible source data changed enough to change the claim.
Warning: 0 shared benchmarks are still too close to call, so the win stays conditional. This compare uses all public sources, with provider-official evidence labeled separately.
Evidence: /artifacts/compare?models=kimi-k2-5-thinking&mode=best-for-this-use-case&preset=coding-copilot
Open in XOpen card preview
Reading guide

How to read this workspace

Who wins most oftenBenchmarks with one clear percentile leader.Missing evidenceBenchmarks where a top model has no visible score.Decision readA public claim tied to an explicit use case and source filter.