Friday, 20 November 2026
Rīga TV

World and Latvian news in one place

TechnologyPublished: 20 November 2026 at 05:05

Google's Gemini 4 Neon Tops the New Models Engineering Index

The Mei index rates language models on practical performance in engineering tasks. Google's Neon model takes first place in its ranking with 68.0 points.

Foto: Hacker News (≥150 p.)

An index called Mei (Models Engineering Index), which calculates the practical performance of language models, or AI systems, on engineering tasks, has been published online. Google's model Neon, named Gemini 4 Neon in the source headline, leads the ranking with 68.0 points. The result matters to anyone choosing models for software development, since the index shows cost alongside quality.

Top of the ranking

Second place, with 65.3 points, goes to Anthropic's Fable 5.2, and third, with 61.5, to OpenAI's 6.2 Astra. Google's 4 Carbon is fourth with 61.4 points, followed by Anthropic's Opus 5.5 with 61.1. Anthropic's Sonnet 5.5 is sixth with 57.8, Google's 4 Hydrogen seventh with 56.9, and OpenAI's 6.2 Sol eighth with 55.9.

Further down are Alibaba's 4-Max (49.0), Moonshot's K3.1 Pro (48.4), OpenAI's 6.1 Terra (47.0) and Meta's Wonder Full (46.4) and Wonder (45.5). The list also includes models from DeepSeek, xAI, Z.ai, Tencent and Xiaomi. OpenAI's 6 Luna is last, in twentieth place, with 34.9 points.

Cost and tokens

The index page compares performance against cost and the number of tokens used. By the cost figure, Neon is 1.6 times and Fable 5.2 is 6.4 times the baseline. The cheapest model on the list is 6 Luna at 0.064 times.

The page also shows results from several benchmarks, including CursorBench, FrontierCode, Code Migration, IOI, ProgramBench, Terminal-Bench and Vibe Code. The source does not explain how the overall score is calculated.

Comments

0/1500

Comments are automatically moderated. No hate, threats, personal data or spam.

Loading comments…

More in this category