Open Source Progress Continues With The Release Of Glm 4 6 In Cline
Source
cline.bot
Author
Cline evaluation and engineering blog
Date
Why it matters
Shows open and closed models nearly tied on diff edits in real AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → usage, with GLM-4.6 priced far below Sonnet 4.5. This directly affects model selection and cost for coding agents.
Key takeaways · AI-distilled
Cline's figures come from millions of diff-edit operations over four months: Claude Sonnet 4 succeeded 95.8% of the time, Sonnet 4.5 96.2% and GLM-4.6 94.9%. Cline treats diff edits as the hardest task because they require precise changes to existing code.
At the list prices Cline cites, GLM-4.6 costs $0.50/$1.75 per million input/output tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → against Sonnet 4.5's $3/$15, a sixth of the input price and under an eighth of the output price. zAI's coding plan offered it at $6/month with 120 prompts per 5-hour cycle.
Cline Discord users described Sonnet 4.5 as needing half the corrections, while GLM-4.6 drew interest for getting close to Sonnet at a fraction of the cost. These are community reports, not measured results.
Cline argues that as coding ability converges, differentiation shifts to ecosystem and tooling, citing AMD running Qwen3 Coder on consumer hardware with 32GB of RAM, while noting open models are not certain to overtake the closed frontier.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
