blog Simon Willison · 1h ago · agents

Anthropic's Opus 4.8 leads model spend as cheaper tools gain share

from English

A Financial Times report indicates that Anthropic's older Opus 4.8 model captured the largest share of billing spend in July, while its newer, more expensive models struggled to attract users compared to cheaper alternatives.

Anthropic's annualized revenue for July reached $65bn, up from $47bn in May, with expectations of Q3 profitability and 6,000 customers spending $100,000 annually or more.
OpenAI's annualized revenue jumped 35% to over $40bn, boosted by the July launch of GPT 5.6.
Ramp AI index data for July shows Opus 4.8 at 28.0% spend share, followed by Sonnet 4.6 (8.3%) and Fable 5 (8.0%), while the newly released Opus 5 held only 3.5%.

The data suggests that cost-sensitive adoption is driving usage toward established or cheaper models, even as newer flagship releases enter the market.

Importance 2/3 Simon Willison Anthropic OpenAI Benchmark results

Read original

arxiv arXiv cs.AI · 11d ago · 17 views

VITA clinical RAG matches or outperforms frontier LLMs on HealthBench

A purpose-built retrieval-augmented generation system named VITA, designed for low- and middle-income settings, was evaluated against general-purpose large language models on the HealthBench benchmark. The study demonstrates that corpus-specific design can maintain competitive performance even as newer frontier models are released.

arxiv arXiv cs.CL · 11d ago · 21 views

Corpus-specific VITA RAG matches or beats frontier LLMs on HealthBench

VITA, a retrieval-augmented generation system designed for low- and middle-income settings, ranks first on the English-language subset of HealthBench, outperforming GPT-5.4, o4-mini, Gemini 3.1 Pro, and Claude Sonnet 4.6.

media MarkTechPost · 37d ago · 27 views

Mistral Vibe for Code leads four coding agents on scaffold-to-PR task

A capability comparison of Mistral Vibe for Code, Claude Code, OpenAI Codex, and Cursor evaluated their performance on a multi-stage workflow involving scaffolding, testing, and pull request generation. Scores were derived from documented features, published benchmarks, and vendor specifications as of July 14, 2026.

media Latent Space · 41d ago · 26 views

OpenAI Codex reaches 7M users, up >10x in 6 months

OpenAI's Codex product has reached 7 million users, representing more than a 10x increase from approximately 550k-700k users at the start of the year. This rapid growth coincides with the July 9 launch of GPT-5.6 Sol and recent infrastructure updates to manage usage.

arxiv arXiv cs.CL · 47d ago · 27 views

RuBench 1.0 introduces Russian-language repository-level agentic coding benchmark

The authors introduce RuBench 1.0, a benchmark of 25 tasks mined from recent fix commits in five live open-source repositories (aiohttp, aiogram, Laravel, NestJS, Fastify). Each task is specified natively in Russian and judged by upstream maintainer regression tests withheld from release.