Matharena is live · free · last verified 22 Oct 2026 · viability monitor
Matharena
Free competition-math LLM leaderboard with Lean-checked proofs and clickable raw model outputs.
If your question is which model actually reasons through elite math, MathArena gives the cleanest free answer, and the clickable raw outputs let you audit the score instead of trusting a summary. The Lean-checked tracks plus the fresh September editions carry real signal. Just don't bring it a coding or chat question — math reasoning is all it scores.
Verdict generatedVerified 8d ago · liveness 62/100 · cite: rightaichoice.com/tools/matharena
- AI researchers measuring mathematical reasoning in frontier LLMs
- Evaluation engineers who need to audit raw model outputs, not just scores
- ML teams picking a math-capable model before a STEM deployment
- Competition-math enthusiasts comparing olympiad-grade performance
- Teams evaluating coding, general knowledge, or conversational ability
- Anyone needing automated benchmarking wired into a CI pipeline or API
- Buyers who need multimodal or diagram-reading evaluation — visual math is deprecated
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip MathArena if you need an automated benchmarking endpoint for your CI pipeline, or a blended multi-domain score covering coding and chat alongside math.
MathArena is a free academic leaderboard from SRI Lab at ETH Zurich and INSAIT — there is no paid tier to weigh against cheaper or pricier evaluation platforms. Its real cost is time: scoring is manual and browser-based, so you budget analyst hours rather than subscription dollars, and you get no automated pipeline or API in return.
In short
Matharena — Free competition-math LLM leaderboard with Lean-checked proofs and clickable raw model outputs. Best for AI researchers measuring mathematical reasoning in frontier LLMs, Evaluation engineers who need to audit raw model outputs, not just scores, ML teams picking a math-capable model before a STEM deployment. Free to use.
What is Matharena?
Matharena — Free competition-math LLM leaderboard with Lean-checked proofs and clickable raw model outputs. Best for AI researchers measuring mathematical reasoning in frontier LLMs, Evaluation engineers who need to audit raw model outputs, not just scores, ML teams picking a math-capable model before a STEM deployment. Free to use.
MathArena is a free public leaderboard that scores frontier language models on hard competition mathematics. Built by SRI Lab at ETH Zurich and INSAIT, it runs three live dataset tracks: ArXivLean (proofs verified with the Lean proof assistant), BrokenArXiv, and ArXivMath. Each track breaks into dated monthly editions plus an Overall view, so you can watch scores move as new problems land. The distinguishing mechanic is auditability. Every leaderboard cell is clickable and opens the raw model output behind the score, letting a reviewer check whether a model genuinely reasoned through a Lean-checked proof or merely hit a lucky final answer. Most aggregate indexes flatten that distinction away. The board is actively maintained. As of the October 16 update the September editions of all three tracks are live at 53 problems each, and the October 17 update added Muse Spark Open results on ArXivMath and BrokenArXiv for June, August, and September. Recent model additions include Gemini 4 Argon, GPT-6.1 Sol (max), Claude-Opus-5 (max), Kimi K3, and Muse Spark 1.1. Mistral Prover (Leanstral 1.5 + K3) sits at 38/46 verified on ArXivLean June 2026. Older views — Kangaroo visual math, AIME and HMMT final-answer contests, USAMO/IMO/Putnam proof sets, Project Euler — are retained but labeled deprecated, so treat the three dataset tracks as the live signal. Against broad multi-domain evaluation platforms, MathArena trades coverage for a narrow, citable read on mathematical reasoning alone.
Key features
- Interactive LLM math benchmark leaderboard
- ArXivLean track with Lean-verified proof checking
- BrokenArXiv track with monthly editions and Overall view
- ArXivMath track spanning 12/2025 through 09/2026
- Click any leaderboard cell to read the raw model output
- Filter leaderboards by competition or dataset month
- Compare models side-by-side across published entries
- September editions of all three tracks with 53 problems each
- Muse Spark Open results on ArXivMath and BrokenArXiv
- Gemini 4 Argon results on ArXivMath and BrokenArXiv
- GPT-6.1 Sol (max) results on ArXivMath and BrokenArXiv
- Mistral Prover (Leanstral 1.5 + K3) at 38/46 verified on ArXivLean June 2026
- Claude-Opus-5 (max), Kimi K3, and Muse Spark 1.1 on the leaderboard
- Citable research blog posts on math evaluation
- Deprecated views retained: Kangaroo visual math, AIME/HMMT, USAMO/IMO/Putnam, Project Euler
Behind the verdict
Most AI leaderboards ask you to take a number on faith. MathArena's argument is that you shouldn't: click any cell and you get the model's actual output, not a summarized score. For anyone who has watched a model land a correct final answer through obviously broken reasoning, that one design choice is the reason to bookmark it. When to pick it: you're choosing a model for a STEM or quantitative workload and want evidence about mathematical reasoning specifically. The ArXivLean track is the interesting one, since proofs get machine-checked rather than judged on the final line. Pair it with the monthly editions if you need to see whether a model has genuinely improved or just memorized older contest sets. The lean setup is also its limit. There's no score export, no API, and nothing you can wire into an evaluation pipeline — this is a browser leaderboard you read. Teams running continuous benchmark suites will find themselves copying numbers by hand. We'd also flag the deprecated sections. Visual math (Kangaroo) and the final-answer and proof-based contest views still load, but the project has marked them deprecated. If multimodal or diagram-reading evaluation is on your list, this is not the place to get it. Compared with broad multi-domain indexes, MathArena is deliberately narrow. That's a feature when math is the thing you're buying for, and a pass when you need coding, general knowledge, or conversational quality in the same table. One practical caveat: model coverage keeps shifting, with entries like Muse Spark Open and Gemini 4 Argon arriving by track rather than everywhere at once. Check that the model you care about is actually scored on the track you're reading before you cite a ranking.
Models under the hood
as of 2026-10-24
Limitations
- MathArena is a free academic leaderboard from SRI Lab at ETH Zurich and INSAIT focused on competition and research-level math.
- The current signal is the three dataset tracks (ArXivLean, BrokenArXiv, ArXivMath); the Visual Math, Final-Answer Comps, Proof-Based Comps, and Project Euler views are all labeled deprecated.
- It only measures mathematical reasoning and Lean proof performance, presenting data via browsable leaderboards rather than a programmatic endpoint.
- No API is described anywhere in the evidence.
as of 2026-10-10
Researching Matharena? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is Matharena still active?
Matharena is live · free · last verified 22 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is Matharena? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What do users say about Matharena?
What people actually say about Matharena — is it worth it?
Our community sentiment research written Jul 3, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
How we verified this
Every listed fact about Matharena— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 22 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified Matharena 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page
- — changed features, pricing tiers, who it suits, summary, who should skip it, our verdict, description
- — changed pricing page, limitations, models, who should skip it, summary, who it suits, our verdict, description, pricing tiers, features
- — changed pricing page, models, pricing tiers, summary, our verdict, features, use cases, who should skip it, setup time, limitations, pricing fit, switching in, switching out, description
- — changed SSL certificate
- — changed who it suits, pricing tiers, features, description, our verdict, summary, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does Matharena cost in 2026?
Matharena: Free to use. Pricing last verified 22 Oct 2026.
Plans compared
Every published Matharena tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free | Free | — | AI researchers, evaluation engineers, and competition-math enthusiasts who need a citable math-reasoning ranking without a budget line — individuals and academic labs alike.Adds: Starting tier, and the only tier: full access to all three live dataset tracks plus the deprecated views, with clickable raw model outputs. |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where Matharena's pricing actually pencils out — and where peers do it cheaper.
MathArena is a free academic leaderboard from SRI Lab at ETH Zurich and INSAIT — there is no paid tier to weigh against cheaper or pricier evaluation platforms. Its real cost is time: scoring is manual and browser-based, so you budget analyst hours rather than subscription dollars, and you get no automated pipeline or API in return.
Has Matharena's pricing changed?
No plan price for Matharena has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick Matharena?
Pick Matharena if you are AI researchers measuring mathematical reasoning in frontier LLMs, Evaluation engineers who need to audit raw model outputs, not just scores, ML teams picking a math-capable model before a STEM deployment, Competition-math enthusiasts comparing olympiad-grade performance.
Look elsewhere if you are Teams evaluating coding, general knowledge, or conversational ability, Anyone needing automated benchmarking wired into a CI pipeline or API, Buyers who need multimodal or diagram-reading evaluation — visual math is deprecated, Users wanting a blended multi-domain index rather than per-track breakdowns.
Skip it if Skip MathArena if you need an automated benchmarking endpoint for your CI pipeline, or a blended multi-domain score covering coding and chat alongside math.
GeneratedReal-world workflow fit
Concrete scenarios for the personas Matharena actually fits — and what changes day-one when you adopt it.
You need a defensible math-reasoning number for a paper's model-comparison table. You open the ArXivLean track, select June 2026, read the ranking, and click the Mistral Prover (Leanstral 1.5 + K3) cell to confirm the 38/46 verified count before writing it into the draft.
Outcome: You cite 82.61% on ArXivLean June 2026 with the underlying model output verified, so a reviewer can reproduce your reasoning instead of taking a percentage on faith.
You want to know whether a model degrades when problem statements contain errors. You load BrokenArXiv, compare a model's Overall view against its newer monthly views through 08/2026, and open individual cells where the score drops sharply.
Outcome: You separate models that hold up under corrupted problem statements from those that only score well on clean prompts, and you have the raw outputs to show why.
Before committing to a math-heavy deployment you run a short bake-off. You filter ArXivMath to the most recent months (through 08/2026), cross-check GPT-6.1 Sol (max) against Claude-Opus-5 (max) and Kimi K3 side by side, and click into the cells where the gap is widest.
Outcome: You pick a model based on where its reasoning actually holds up on olympiad-grade problems, rather than on a marketing chart or a deprecated benchmark.
Use cases
- Check how a frontier model performs on ArXivLean June 2026 before citing it in a paper
- Audit a proof-based score by opening the raw model output behind a leaderboard cell
- Use BrokenArXiv to stress-test a model's robustness to erroneous problem statements
- Track how Mistral Prover (Leanstral 1.5 + K3) compares to Claude-Opus-5 (max) on the same track
- Pull a free, citable math-reasoning ranking to justify a model choice in a STEM deployment
- Compare monthly ArXivMath snapshots 12/2025 through 08/2026 to see whether a model is improving or plateauing
Setup time & first value
How long it actually takes to get something useful out of Matharena — broken out by persona, not the marketing-page minute.
Researchers and evaluation engineers: a few minutes to first value — the site loads straight into the leaderboard, you select a competition from the dropdown, and you are reading rankings immediately, with no account required. Competition-math enthusiasts: roughly the same, though plan a little extra time to learn which tracks are live (ArXivLean, BrokenArXiv, ArXivMath) versus the deprecated
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “Matharena”, and we withheld 6: 6 could not be judged, because “Matharena” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Matharena.
Official links
What are the best Matharena alternatives?
The closest alternatives to Matharena we list are Opencompass, Triall and Goodfire. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Opencompass | Free | Free | — |
| Triall | Freemium | Pricing not published | — |
| Goodfire | Paid | Pricing not published | — |
Alternatives to Matharena
View all
Opencompass
OpenCompass (司南) benchmarks LLMs, VLMs, and AI4S models on dated, dimension-by-dimension leaderboards
Tools that pair well with Matharena
Common stack mates teams adopt alongside Matharena, with the specific reason each pairing earns its keep.
Opencompass
OpenCompass (司南) benchmarks LLMs, VLMs, and AI4S models on dated, dimension-by-dimension leaderboards
Triall
Put a hard question to three frontier models — they answer blind, cross-examine each other, and hand back the one answer that survived, with a verdict.
Goodfire
Goodfire's Silico is an AI interpretability agent for reverse-engineering, debugging, and steering model internals.
Matharena vs its closest rival
Matharena's closest head-to-head is with Surge Ai — For researchers benchmarking open-source LLMs on elite math, Matharena is the free, community-driven choice with immediate access to competition data. Surge AI is the expert human-in-the-loop platform for frontier labs that need rigorous, domain-expert evaluations for RLHF and safety — at a higher cost but with deepertailoring. Choose based on whether you need automated math benchmarks or nuanced human feedback. Verdict generated
Head-to-head comparisons
Matharena vs Surge Ai
For researchers benchmarking open-source LLMs on elite math, Matharena is the free, community-driven choice with immediate access to competition data. Surge AI is the expert human-in-the-loop platform for frontier labs that need rigorous, domain-expert evaluations for RLHF and safety — at a higher cost but with deepertailoring. Choose based on whether you need automated math benchmarks or nuanced human feedback.
GeneratedMatharena vs Praktika
Praktika and MathArena serve entirely different needs. Praktika is ideal for intermediate language learners seeking conversational fluency via AI tutors, while MathArena is a specialized benchmarking tool for evaluating LLMs on rigorous math problems. Choose based on your goal: practicing English or testing AI reasoning. There's no overlap.
GeneratedFrequently asked questions
Categories
Topics
Used Matharena? Help shape our editorial sentiment research.