Is LMArena profitable in 2026?

LMArena has not disclosed profitability. It reported an annualized revenue run rate of about 100 million dollars by June or July 2026, per its own company statement, up from 30 million dollars in December 2025, but no profit figures or audited financials have been published.

How much is LMArena worth?

LMArena, now rebranded Arena, was valued at 1.7 billion dollars post money after a 150 million dollar Series A round announced January 6, 2026, led by Felicis and UC Investments, roughly triple its 600 million dollar seed valuation from May 2025.

Could AI replace LMArena?

Automated LLM as judge scoring already aligns with human preference 80 to 85 percent of the time as of May 2026, which could reduce reliance on LMArena's human vote model over time, though its 82 million plus vote dataset and neutral referee brand remain hard to replicate quickly.

Who are LMArena's biggest competitors?

Surge AI launched a rival Antidote Leaderboard in May 2026 explicitly as an Arena alternative, while Scale AI runs its own SEAL leaderboards and Hugging Face and Artificial Analysis offer competing model comparison platforms.

Is LMArena's benchmark trustworthy?

LMArena's methodology has faced sustained challenges, including an April 2025 Cohere, Stanford, MIT and Ai2 study accusing it of helping labs game its benchmark, and a July 2026 report that Meta manipulated its Llama 4 Maverick score, which dropped from 2nd to 32nd place.

What is LMArena's biggest business risk?

LMArena's revenue is concentrated among a small number of frontier lab customers like OpenAI, Google DeepMind, Anthropic and xAI, who are also potential disintermediators if they build in house evaluation tools, and its 100 million dollar run rate figure from June 2026 is self reported and unaudited.

United States · United States

LMArena

Arena (formerly LMArena) runs public leaderboards where people vote on AI model outputs.

Protection under test

Assessed 29 July 2026 · Evidence as of 29 July 2026

Something worth protecting exists and is being tested right now.

On the recordSince 29 July 2026 · Checkable from Q4 2026 · Every call on the record Checkpoint: Next funding round where $100M ARR claim faces investor scrutiny and possible audit

The call

Arena sits on a large pile of human preference data and fast revenue growth, but its neutrality claim is under public challenge and its biggest customers are also its biggest potential rivals.

LMArena is a platform that tests AI systems by collecting human votes on which model performs best, then publishes leaderboards that labs and companies use to benchmark themselves. The hard part is staying neutral and trustworthy, which OpenAI, Google, and Anthropic cannot do for themselves because they would be grading their own work. But LMArena's trust is being attacked: researchers published studies questioning its methods, a rival called Antidote launched a competing leaderboard, and there was a gaming scandal involving Meta. Money comes from a handful of big AI labs, so if one leaves or builds its own testing system, LMArena loses a huge chunk of revenue. The one thing that matters most: do customers keep paying after these trust scandals, or do they walk away?

Why

AI pressure: Automated LLM-as-judge scoring systems built by frontier labs. Machine judges now match human votes 80-85% of the time, chipping at the need for Arena's crowdsourced voting model. Could AI do the core job? Only partly.

Bigger players: OpenAI, Google DeepMind, Anthropic, xAI. These labs fund Arena's evaluations today but are also building internal eval tools that could cut their reliance on it. Could a giant copy it? Probably.

What protects it. 82M+ human votes and a neutral-referee reputation that no single lab can credibly replicate on its own.

What would change this call

A frontier lab (OpenAI, Google DeepMind, Anthropic, xAI) publicly drops Arena as pre-release tuning partner. Would show revenue-concentrated customers are disintermediating Arena's core moat.
Surge AI's Antidote or Scale SEAL announces a named enterprise account switching from Arena. Would show the neutrality moat is losing to a direct rival's leaderboard.
Arena's next funding round discloses audited revenue figures matching the self-reported $100M run-rate. Would confirm or undercut trust in Arena's unaudited growth claims.

Key numbers

$100M annualized revenue run-rate reached (2026-06-29)
$1.7B valuation after Series A (2026-01-06)
5 million monthly users across 150 countries (2026-01-06)

Named risks

Revenue growth from 30 million to 100 million dollar run rate is self reported and unaudited. Severity: high
Methodology credibility attacked by Cohere Stanford MIT study and Surge AI campaign. Severity: high
Revenue concentrated among a few frontier lab customers who could build in house evaluation. Severity: high
Automated LLM as judge scoring could erode reliance on human crowdsourced votes. Severity: medium
Post rebrand confusion as many still search LMArena instead of Arena. Severity: low

Rivals, as read in July 2026

Rival	Direction	Evidence
Surge AI	contesting	Launched Antidote Leaderboard May 2026 as direct Arena alternative
Scale AI	contesting	Runs competing SEAL leaderboards for model evaluation
Meta	losing to	Meta gamed its Llama 4 Maverick Arena score, dropping from 2nd to 32nd, reported July 2026
Hugging Face	contesting	Open, self-hostable model hub cited as alternative for model comparison
Artificial Analysis	gaining on	Independent benchmarking site increasingly cited alongside Arena in mid 2026

Questions people ask

Is LMArena profitable in 2026?: LMArena has not disclosed profitability. It reported an annualized revenue run rate of about 100 million dollars by June or July 2026, per its own company statement, up from 30 million dollars in December 2025, but no profit figures or audited financials have been published.
How much is LMArena worth?: LMArena, now rebranded Arena, was valued at 1.7 billion dollars post money after a 150 million dollar Series A round announced January 6, 2026, led by Felicis and UC Investments, roughly triple its 600 million dollar seed valuation from May 2025.
Could AI replace LMArena?: Automated LLM as judge scoring already aligns with human preference 80 to 85 percent of the time as of May 2026, which could reduce reliance on LMArena's human vote model over time, though its 82 million plus vote dataset and neutral referee brand remain hard to replicate quickly.
Who are LMArena's biggest competitors?: Surge AI launched a rival Antidote Leaderboard in May 2026 explicitly as an Arena alternative, while Scale AI runs its own SEAL leaderboards and Hugging Face and Artificial Analysis offer competing model comparison platforms.
Is LMArena's benchmark trustworthy?: LMArena's methodology has faced sustained challenges, including an April 2025 Cohere, Stanford, MIT and Ai2 study accusing it of helping labs game its benchmark, and a July 2026 report that Meta manipulated its Llama 4 Maverick score, which dropped from 2nd to 32nd place.
What is LMArena's biggest business risk?: LMArena's revenue is concentrated among a small number of frontier lab customers like OpenAI, Google DeepMind, Anthropic and xAI, who are also potential disintermediators if they build in house evaluation tools, and its 100 million dollar run rate figure from June 2026 is self reported and unaudited.

Timeline

2026-07-02Code Arena expands to fullstack AI evaluation, ranking 104 models
2026-06-29Company reports reaching 100 million dollar annualized revenue run rate
2026-06-04Agent Mode launched, extending evaluation to multi step agentic workflows
2026-05-20Surge AI launches Antidote Leaderboard as explicit Arena rival
2026-02-09Max model router using community vote data launched
2026-01-28Company rebrands from LMArena to Arena, moves domain to arena.ai
2026-01-21Video Arena support launched
2026-01-06150 million dollar Series A closes at 1.7 billion dollar valuation
2025-09-16Paid product AI Evaluations launches for enterprises and labs
2025-05-21100 million dollar seed round closes at 600 million dollar valuation

Revision history

29 July 2026

First assessed. Protection under test. The text above is fixed as of this date and is not edited afterwards.

In the same band

Ford Motor Whirlpool Solventum