Agnes 2.5 Flash: The New Free Coding Model, on OpenClaw Launch

What Is Agnes 2.5 Flash?

Agnes 2.5 Flash is the successor to agnes-2.0-flash, the free multimodal model already available on OpenClaw Launch through the Agnes API key integration. Agnes AI positions 2.5 Flash as the model built for everyday developer use — coding, agent workflows, and complex development scenarios — and reports significant gains over 2.0 in four areas:

• Code comprehension across large, unfamiliar codebases
• Bug detection and engineering fixes in existing projects
• Multi-step, cross-file task execution
• Complex reasoning

In Agnes AI’s own demonstrations, 2.5 Flash detects and fixes bugs in existing codebases, builds full applications from scratch, and makes coordinated cross-file edits — with results the company describes as competitive with Claude Opus 4.7. It has a large 512K token context window, supports up to 65,536 output tokens, and speaks the standard OpenAI-compatible chat format, so nothing in your OpenClaw setup needs to change to use it.

Is It Really Free?

Agnes currently lists Agnes 2.5 Flash at $0 per million input and output tokens. Access is still governed by your Agnes account’s entitlements and service limits. Agnes also offers the separately billed Agnes 2.5 Pro Alpha for heavier reasoning workloads.

Availability status (verified 27 July 2026)

Both 2.5-generation chat models are now served by the OpenAI-compatible Agnes API:

• agnes-2.5-flash — live and suggested. OpenClaw Launch lists it first in the Agnes model picker and uses it for new Agnes-funded deployments.
• agnes-2.5-pro-alpha — live on the API now. The paid reasoning model remains an explicit opt-in and is never selected automatically.

Existing instances are not silently switched. If your saved primary is agnes-2.0-flash, it stays there until you choose 2.5 Flash; 2.0 remains available as a compatibility fallback.

Agnes 2.5 Pro Alpha: What’s Live Today

agnes-2.5-pro-alpha is Agnes’s billed 2.5-generation reasoning model. It is offered in the model picker, labelled Agnes 2.5 Pro Alpha (Paid). Three things to know before you switch:

• It is not free. Unlike every other Agnes model on OpenClaw Launch, Pro Alpha bills your own Agnes account balance per request. If your balance runs short, Agnes rejects the call with an insufficient-quota error rather than falling back. Top up on the Agnes platform before you rely on it.
• It is a reasoning model. It thinks before it answers, and that thinking consumes output tokens — the bulk of them on short replies. We declare a much larger output budget for it than for the Flash models so you get a real answer instead of an empty one, but expect higher token use per turn than 2.0 Flash for the same question.
• It is an alpha. The name is Agnes’s, not ours. Treat it as a model to evaluate rather than one to leave running unattended on a production bot, and keep a free Flash model as your fallback.

It accepts images as well as text, so multimodal prompts work the same way they do on 2.0 Flash. Because it is paid, it is never selected for you: it is never a default, and it is never applied to an instance automatically. You get it only by choosing it.

How to Use Agnes 2.5 Flash on OpenClaw Launch

The setup uses the same Agnes key that already unlocks 2.0 Flash. Once you have it saved, switching to 2.5 Flash takes one click.

1. Get a free Agnes key — Sign up at platform.agnes-ai.com (no card required) and create an API key. Agnes keys start with sk-. Full walkthrough: the Agnes API key setup guide.
2. Save it on OpenClaw Launch — Open the API Keys page, find the Agnes card, paste your sk- key, and click Save. We validate it against the Agnes API before storing, and it is encrypted at rest.
3. Set your chat model — In the Agnes card, click “Set as primary model” and pick your model. The suggested option is agnes-2.5-flash. agnes-2.0-flash remains available as a fallback, and agnes-2.5-pro-alpha appears in the same dropdown marked (Paid), billed to your Agnes balance if you choose it. Applying takes effect on your running instances immediately — no redeploy.

This works for both frameworks. On OpenClaw, the model routes directly through your declared Agnes provider; on Hermes Agent, the same key is wired into the custom-provider slot at model-switch time. Either way you use the Agnes gateway with your own key; Flash is currently $0, while Pro Alpha is paid.

Should You Switch From 2.0 to 2.5?

For 2.5 Flash: yes for coding and agent work — that is exactly the workload it was built to improve, and it is the same price ($0). If your bot mostly does lightweight chat, 2.0 Flash remains perfectly capable and stays available. Because switching is a one-click model change with no redeploy and no cost, the low-risk move is to try it and keep it if you like the results.

For 2.5 Pro Alpha, the calculus is different because it costs money. It is worth trying if you are hitting the ceiling of what 2.0 Flash can do on hard multi-step or debugging work and need its reasoning behavior. It is not worth switching a chatty, high-volume bot onto: reasoning tokens on every turn add up, and the free model handles that workload fine. A sensible pattern is to keep 2.5 Flash as your everyday model and switch to Pro Alpha for a specific hard task, then switch back.

Troubleshooting

I don’t see agnes-2.5-flash in my model picker yet

• Refresh the API Keys page. Agnes 2.5 Flash is now in the platform catalog and should appear as the suggested option.
• Make sure your Agnes key is saved and confirmed on the API Keys page — the Agnes section only shows once a valid key is on file.

Model Not Found (404)

• Use the exact id agnes-2.5-flash. If a saved key still gets a 404, re-save it on the API Keys page and contact support with the time of the failed request; the model is currently served by the Agnes API.

401 Unauthorized

• Confirm the key starts with sk- and has no stray spaces or newlines. Regenerate it on the Agnes dashboard and re-save if needed.

Insufficient quota on Agnes 2.5 Pro Alpha

• Pro Alpha is billed to your own Agnes balance, and Agnes checks the balance before running the request — so a long prompt can be rejected outright when your balance is low, even though shorter prompts still work. Top up on the Agnes platform, or switch back to free agnes-2.5-flash.
• The free Agnes Flash models are not billed like Pro Alpha — if Pro Alpha is erroring on quota, switch to 2.5 Flash or the 2.0 fallback.

Agnes 2.5 Pro Alpha replies look empty or truncated

• That is the reasoning behaviour: the model spends output tokens thinking before it writes, so a tight output limit can be consumed entirely by reasoning, leaving no visible answer. Instances configured through the model picker already get a reasoning-sized output budget. If you have hand-edited a small max-tokens value in your own config, raise it.

Frequently Asked Questions

What is Agnes 2.5 Flash?

Agnes 2.5 Flash is Agnes AI’s newest free flagship model, the successor to Agnes 2.0 Flash. It delivers significant improvements in code comprehension, engineering fixes, multi-step task execution, and complex reasoning, with performance Agnes describes as comparable to Claude Opus 4.7. Agnes currently prices it at $0, with a 512K token context window.

Is Agnes 2.5 Flash free?

Agnes currently lists Agnes 2.5 Flash at $0 per million input and output tokens. Account entitlements and service limits still apply. The separate Agnes 2.5 Pro Alpha model is billed to your Agnes balance.

Can I use Agnes 2.5 Flash on OpenClaw Launch right now?

Yes. Agnes 2.5 Flash is live through the Agnes API and appears as the suggested Agnes model on OpenClaw Launch. Selecting it updates running instances without a redeploy. Existing instances saved on agnes-2.0-flash remain there until you switch them.

What is Agnes 2.5 Pro Alpha, and is it free?

It is Agnes’s premium 2.5-generation reasoning model. It is not free: it bills your own Agnes account balance per request, and Agnes rejects calls when your balance is short. It shows in the OpenClaw Launch model picker as “Agnes 2.5 Pro Alpha (Paid)”, and it is never applied to your instance automatically — the suggested default stays a free model.

Why does Agnes 2.5 Pro Alpha use more tokens than 2.0 Flash?

Because it is a reasoning model: it produces internal thinking before its visible answer, and that thinking is billed as output. On short questions the reasoning can account for the large majority of the tokens used. That is the trade for stronger multi-step and debugging performance — for routine chat, free 2.5 Flash is the more sensible model.

How is Agnes 2.5 Flash different from 2.0 Flash?

2.5 Flash is a substantial upgrade focused on development work: better code understanding, stronger bug-fixing on existing projects, more reliable multi-step and cross-file execution, and improved complex reasoning. Both are free and share the same OpenAI-compatible API. 2.5 Flash expands the context window to 512K, while 2.0 remains available as a compatibility fallback.

Does it work with Hermes Agent too?

Yes. The same free Agnes key works on both OpenClaw and Hermes Agent instances. Save your key once and pick the Agnes model from the picker on either framework.

What's Next?

• Agnes API key setup on OpenClaw — The full step-by-step for getting and saving your free Agnes key
• Agnes on Hermes Agent — Same free key, set up for Hermes Agent instances
• AI coding plans compared — How free models like Agnes stack up against paid coding plans
• API Keys page — Add or manage all your provider keys in one place