Welcome back to Agentic Coding Weekly. Here are the updates on agentic coding tools, models, and workflows for the week of Jun 7-13, 2026.
Executive Summary:
Anthropic released Claude Fable 5 and Mythos 5. Scored 80.3% on SWE-bench Pro and 88.0% on Terminal-Bench 2.1.
Access did not last long. On June 12, Anthropic pulled access after a US government request to suspend access for foreign nationals.
Moonshot AI released a coding model, Kimi K2.7 Code. Did not publish agentic coding benchmark numbers.
Zai released GLM 5.2 on its coding plan. Weights coming in a week. No benchmarks yet.
Xiaomi launched MiMo V2.5 Pro UltraSpeed and released MiMo Code, a fork of OpenCode.
Worth reading: a short post on how to show consideration for teammates while sending AI generated content, a career reflection from a senior engineer, and an HN thread on personal tools built with AI.
Meanwhile, a giant star may have destroyed itself in one of the universe's rarest explosions. No, I'm not talking about Fable 5.
1. Tooling and Model Updates
Claude Fable 5 and Mythos 5
Anthropic finally released Mythos class model publicly last week on June 9th. Mythos 5 is same model as Fable 5 but with less safeguards and only available to select customers.
On agentic coding it scores 80.3% on SWE-bench Pro, ahead of the Mythos Preview model (77.8%), Opus 4.8 (69.2%), GPT-5.5 (58.6%), and Gemini 3.1 Pro (54.2%). On Terminal-Bench 2.1, it scores 88.0% against Opus 4.8's 82.7%, GPT-5.5's 83.4%, and Gemini 3.1 Pro's 70.7%. Pricing is $10 / $50 per million input / output tokens, less than half the Mythos Preview rate.
On June 12th, US government asked anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national. Anthropic pulled access from everyone and as of today, it's not available to any customer.
Quick Updates
Moonshot AI released Kimi K2.7 Code, a coding model built upon Kimi K2.6. Although it's a coding model, they didn't publish the benchmark scores for agentic coding benchmarks like SWE-bench Pro and Terminal Bench 2.1. Pricing is $0.95 / $4 per million input / output tokens.
Zai released GLM 5.2 but it's available only on their coding plan so far. They will release the weights in a week or so. No benchmark results yet.
Xiaomi is running a limited time experiment with MiMo V2.5 Pro UltraSpeed. This model variant that hits 1000 TPS, about 10x faster than MiMo V2.5 Pro. It's also 3x more expensive and costs $1.305 / $2.61 per million input / output tokens.
Xiaomi also launched their own coding agent, MiMo Code. It's a fork of OpenCode and they have added some new features on top of it.
2. Community Picks
If You are Asking for Human Attention, Demonstrate Human Effort
Concise and self-explanatory. Bookmark it to send as a response when somebody asks you to review their slop.
LLMs Are Eroding My Software Engineering Career and I Don’t Know What to Do
Author is a software engineer with 10 years of experience. Argues that LLMs and coding agents have already commoditized domain knowledge and debugging distributed systems which are two of his core competencies. Coding agents still haven't matched his capability on code quality and architecture, but code quality is becoming less important in the industry. Reflects on what this means for employability going forward. Read the post.
Ask HN: What are tools you have made for yourself since the advent of AI?
Check the HN thread. What kind of things have you been building?
That’s it for this week. I write this weekly on Mondays. If this was useful, subscribe below:
— Prashant
