SemiAnalysis: GLM and Kimi now do the work that built Anthropic's $65B ARR
iScienceLuvr · x · 2026-08-22
SemiAnalysis publishes "Are Open Models Catching Up?" comparing open vs closed models across frontier, reasoning, and agentic eras:
- The "DeepSeek moment" (Jan 2025) was hype — nobody used R1 for economically valuable work. But GLM 5.3 and Kimi K3 now genuinely handle the coding and agentic tasks that rocketed Anthropic to $65B+ ARR.
- The token battle is heating up: Fireworks alone processes 40T+ tokens/day — 2x OpenAI's API volume at end of March; competition extends beyond the OpenAI–Anthropic duopoly.
- Open success fuels FUD: if open models approach the closed frontier at a fraction of the cost, will the model layer commoditize and destroy frontier-lab margins?
- Methodologically, no single benchmark spans history — each benchmark is a product of its era and saturates; LLM history splits into early scaling, reasoning, and agentic eras.
Full post is paywalled; these are the public takeaways.
Related event: SemiAnalysis: Open-Source Models Are Closing the Gap with Closed Frontiers(2 posts)→
More from AGI Musings
- Minerva brings AI to a 3,000-person village and acquires accounting firms across the US — ycombinator · 2026-10-16
- Interpretability is too costly for everything, but great for post-hoc forensics — jessi_cata · 2026-10-16
- Neuron labels are like code comments: debate on interpretability vs CoT — jessi_cata · 2026-10-16
- Jensen Huang: UK Must Fix Energy Prices to Seize Its AI 'Goldilocks Moment' — KeeganMcB · 2026-10-16
- SF Vibes Are Electric as Engineers Move From Automation Anxiety to Awe — zack_overflow · 2026-10-16
- Neel Nanda: interpretability is useful but nowhere near reliable enough to ensure AI safety — aryaman2020 · 2026-10-16