Category intelligence
86 current items analyzed and ranked.
Executive synthesis
Primary evidence
xAI released Grok 4.6, focused on long-running tasks and priced below competing frontier models. The release comes one day after Google's Gemini 3.7 Flash, intensifying the same-week pricing and capability competition.
By Matthias Bastian
Google released Gemini 3.7 Flash, three weeks after 3.6 Flash, claiming it beats Claude Sonnet 5 and GPT-5.6 Terra on coding benchmarks at roughly half the prior Flash price. The model ships at $0.75 per million input tokens, signaling aggressive pricing in the workhorse tier.
Google introduced Gemini 3.7 Flash, billed as the most intelligent workhorse model yet for coding and agents, released just three weeks after Gemini 3.6 Flash and incorporating developer feedback plus algorithmic improvements.
By Ryan Whitwam
Google rolled out Gemini 3.7 Flash just three weeks after 3.6 Flash, positioning it as a workhorse with notable coding gains (FrontierCode 1.1: 34.4 to 43.6; DeepSWE v1.1: 49 to 65.3) and a lower introductory price to counter cheaper competitors.
By Julie Bort
Databricks raised $5B at a $190B valuation after investor demand pushed the round well past the $1B originally targeted, with CEO Ali Ghodsi citing AI capex intensity.
By Asif Razzaq
MarkTechPost details Google Gemini 3.7 Flash, an algorithmic refinement of 3.6 Flash with a 1M-token context window, 64K output, customizable thinking budgets, and a March 2026 knowledge cutoff. Improvements concentrate in software engineering, document-heavy knowledge work, and web development, at $0.75/$3.75 per million input/output tokens.
By George Hammond, Financial Times
Continuing our coverage from yesterday, Anthropic backers told the FT they expect a fall IPO at a $2T+ valuation, which would eclipse SpaceX and rank as the largest-ever IPO, contingent on rapid revenue growth.
By Matthias Bastian
According to Ramp data, Anthropic's Claude Fable 5 accounts for only about 6% of Anthropic tokens sold, suggesting U.S. enterprises are reluctant to pay the premium price. The analysis frames this as evidence that corporate AI spending may have plateaued absent clear everyday productivity gains.
By Asif Razzaq
Liquid AI released LFM2.5-VL-3B, a 3.1B-parameter vision-language model for on-device deployment that reads digital screens, grounds objects to coordinates, parses documents and charts, and calls tools from text or images. It ships in native, GGUF, ONNX, and MLX formats and averages 69.4 across 28 vision benchmarks, matching 4.7B-class models.
DeepSeek released DeepSeek Harness (dsh), an open-source agent harness built around an everything-is-a-plugin architecture powered by its Cordis runtime. It ships as a developer preview via npm and source builds, with a Web UI on localhost.
By Rebecca Bellan
Anthropic research finds that AI agents given the same task can clash, collude, and coordinate in unexpected ways, suggesting current safety evaluations may miss multi-agent risks.
European vibe-coding startup Lovable has reached a $13.3 billion valuation, reflecting rapid international investor interest in the AI coding-application space. The figure underscores the premium investors are placing on AI-native developer tools.
