
Local LLMs in 2026: What Actually Runs on a Laptop Now
What a freelancer can actually run locally in August 2026: Qwen 3.8-27B on a 32 GB Mac, why full DeepSeek V4-Flash is not a laptop job, and a one-weekend experiment.
Topic desk / 04
Agents, coding assistants, local models, and harness work. Original topic labels remain visible inside this desk.
Start here

What a freelancer can actually run locally in August 2026: Qwen 3.8-27B on a 32 GB Mac, why full DeepSeek V4-Flash is not a laptop job, and a one-weekend experiment.
As of August 2026, prompt skill is baseline. A freelancer can build a production-shaped agent with free or self-hosted infrastructure components; model inference is a separate cost or local choice.
A field test of running Grok, Claude, and Codex as one content pipeline on DeepSeek Harness—roles, real commands, subscription-quota cost, and when the extra steps are worth it.
Decision paths
Topic ledger
No guides match this format. Clear filters
What a freelancer can actually run locally in August 2026: Qwen 3.8-27B on a 32 GB Mac, why full DeepSeek V4-Flash is not a laptop job, and a one-weekend experiment.
As of August 2026, prompt skill is baseline. A freelancer can build a production-shaped agent with free or self-hosted infrastructure components; model inference is a separate cost or local choice.
A field test of running Grok, Claude, and Codex as one content pipeline on DeepSeek Harness—roles, real commands, subscription-quota cost, and when the extra steps are worth it.
A practical test for deciding whether agentic coding tools earn their keep on personal, unpaid projects—and where they quietly cost you.
A practical guide to choosing between autocomplete, chat-based, and agentic coding assistants based on your work, risk tolerance, and budget.
Other desks
Changed terms and deadlines worth checking. No promo blasts.
Subscriptions are not open yet. New guides are available in the library.
Browse guides