Gemini 4 Neon is the new #1 on the Models Engineering Index

Mei, short for the Models Engineering Index, is a public ranking on the git-aarvin.github.io site. It scores language models on practical engineering work and shows cost and token use next to performance. In the current reading, Google's Gemini 4 Neon ranks first with 68.0, ahead of Anthropic's Claude Fable 5.2 at 65.3 and OpenAI's GPT-6.2 Astra at 61.5, according to a secondary write-up that says the data comes from Cognition, Cursor, the Laude Institute and Vals AI. The index's own chart text, however, lists Neon at 69.5, Fable at 65.8 and Astra at 60.5. The published figures therefore do not match across copies, and the material reviewed does not spell out the scoring formula.
Google released Neon on November 18, 2026 as the middle model of its Gemini 4 family. It is priced at $3 per million input tokens and $15 per million output tokens, with a 1 million token context window. Google says Neon ranks third on the Artificial Analysis Intelligence Index with a score of 61 at its highest thinking setting, and first on that firm's Coding Agent Index at 58.6.
Cost is a weaker point on some measures. At its highest setting, Neon used 283 million output tokens across the Artificial Analysis index, which that firm described as notably slow and very verbose.
Why it matters
For teams building coding agents or engineering tools, the ranking makes Neon a serious candidate for the model layer, especially where cost per task matters. Rankings still disagree on order, and results depend on which models each board includes. Neon trails Fable 5.2 on several boards, so a single leaderboard position is a thin basis for locking in a default model. Teams should run their own evaluations on their own repositories and workflows before switching.
For investors and operators, the result adds pressure on companies that resell or wrap one frontier model, because suppliers now ship competitive models on short cycles. Google has released four Gemini 4 models, with Neon as the fourth within roughly seven weeks. Per-task cost and token consumption can shift gross margins as much as benchmark placement, so founders should track unit economics per workflow. Mei's methodology is not fully visible in the material reviewed, so it should be treated as an early signal.
Key points
- 1Mei places Gemini 4 Neon first at 68.0, ahead of Claude Fable 5.2 at 65.3, according to a secondary report.
- 2Google released Neon on November 18, 2026, priced at $3 per million input tokens and $15 per million output.
- 3Google reports Neon at 58.6 on the Artificial Analysis Coding Agent Index, the top score on that board.
- 4Neon scores 61 on the Artificial Analysis Intelligence Index, behind Fable 5.2 at 65 and GPT-6.2 Astra at 62.