Hermes opens 7 free models, run auxiliary tasks with zero API Key

📡 AI News 2026-09-10 3 min read

Hermes opens up 7 free models, run auxiliary tasks with zero API Key. Hermes recently integrated OpenCode's free model pool. 7 models, not a single cent, no API Key needed, no account registration needed, ready to use right away. People who frequently run Agents can probably feel that auxiliary tasks consume quite a few tokens. Context compression, title generation

💡 What You Will Learn

Hermes opens up 7 free models, run auxiliary tasks with zero API Key. Hermes recently integrated OpenCode's free model pool. 7 models, not a single cent, no API Key needed, no account registration nee

📜 Table of Contents

Hermes Opens Up 7 Free Models, Run Auxiliary Tasks with Zero API Key

Hermes recently integrated OpenCode's free model pool. 7 models, not a single cent, no API Key needed, no account registration, ready to use out of the box.

Anyone who runs Agents regularly has probably noticed that auxiliary tasks consume a fair number of tokens. Context compression, title generation, webpage summarization — every time these go through a paid model, the end-of-month bill doesn't look great. Now these jobs can be offloaded to the free pool.

What's in the Free Model Pool

Currently OpenCode's free tier has 7 models.

Model Type
deepseek-v4-flash-free General conversation
muse-spark-1.2-contributor-free General conversation
mimo-v2.5-free General conversation
ling-3.0-flash-fin-free Finance domain
nemotron-3-ultra-free General conversation
nemotron-3.5-lightning-free General conversation
laguna-s-2.1-free General conversation

All carry the -free suffix, go through OpenCode's free channel, send requests anonymously, and don't consume any paid quota.

This list will change. OpenCode periodically rotates the models in the free tier — new ones come online, old ones retire. Hermes now reads OpenCode's live catalog directly, no longer relying on a locally frozen list, so models entering and leaving sync automatically.

The Most Worthwhile Use Cases

Using the 7 free models as your main workhorse isn't realistic — both capability and speed have limits. But for auxiliary tasks, their value shows up.

Hermes supports assigning different models to different background tasks.

These jobs don't inherently need your strongest model. Context compression is just condensing a long conversation into a summary, title generation is just naming a session, webpage summarization is just distilling page content into a few paragraphs. Using a paid model for these is like using a cannon to swat a mosquito.

Switch auxiliary tasks to the free pool, and let paid models focus their energy on conversations and tool calls that need quality. Spend money where it counts.

Model List Auto-Refresh

Hermes now checks OpenCode's free model catalog in real time, no longer using a local static list.

If OpenCode pulls a model from the free tier, it automatically disappears on Hermes's side. When a new model is added to the free tier, Hermes discovers it automatically. The whole process doesn't require waiting for a Hermes release update.

The free list will change — that's by design. OpenCode is adjusting its free strategy, and Hermes just follows along.

How to Use

There are two ways to switch to the free model pool.

Type hermes model on the command line to enter the interactive model selector, and choose OpenCode Free.

In a session, type /model free to switch directly.

After switching, the main conversation model becomes a free model. If you want to switch only auxiliary tasks to the free pool while keeping a paid model for the main model, you can separately configure the auxiliary section in config.yaml, pointing tasks like compression and title generation to opencode-free.

Practical Advice

If you're already using Hermes, try switching all auxiliary tasks to the free pool and observe the cost change over a round of conversation. Tasks like compression and title generation don't demand much model capability, and free models are fully up to the job. Keep your usual paid model for the main conversation, and leave the free quota for background chores.

Related Articles
2026-08-06
Goldman Sachs: Money Is Exiting the Big Seven Tech Giants, Semiconductors Are the Next Big Bet
2026-07-17
Comfy Desktop Launched: AI Image Generation for Everyone, No Terminal Required
2026-07-18
Win11 K2 Plan Reveals: Right-Click Menu, Search, AI Management — 4 Pain Points Getting Fixed
2026-10-22
Macs are quite capable at running large models, but fall apart when it comes to video: measured 30 times slower
2026-09-21
A laptop stuffed with 192GB of memory: AMD wants to let large models live directly inside the laptop
2026-07-17
Win11 Bulk Delete Gets 30%+ Faster: Microsoft Finally Modernizes File Explorer

Written by our editorial team; tools listed here are tested or verified against public sources. Links point to official sites or GitHub repos for reference only — no paid placements.

💬 Comments (0)

No comments yet. Be the first!

Login to comment