Today we're launching Eragon Flux, model routing for enterprise agents. Flux breaks a task into steps and runs each one on the cheapest model that can get the job done, reserving frontier models for the hardest steps. Cost per task keeps falling as Flux learns. 🧵
25
4
36
1,855
Today we're launching Eragon Flux, model routing for enterprise agents. Flux breaks a task into steps and runs each one on the cheapest model that can get the job done, reserving frontier models for the hardest steps. Cost per task keeps falling as Flux learns. 🧵
25
4
36
1,855
4/ In our deployments, we observe that quality often improves as costs fall. Subagents scoped to a single step see less irrelevant context and make fewer errors. As Eragon keeps running a task, Flux evaluates cheaper models on it and shifts traffic only when quality holds.
1
1
77
Flux is in production at leading companies like Deel, Slash and Prodware, for engineering and business teams alike. With an Eragon key, engineers use model routing in Claude Code, Cursor, and Codex. GTM, finance and operations teams automatically get routing in the Eragon app.
Read more: eragon.ai/blog/introducing-e…
1
2
95
GPT-6.1 Terra is now available in Eragon.
Strong performance at a fraction of the cost, ready to use today.
GPT-6.1 Terra: delivering fast and capable computer use at a lower price.
GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
9
3
12
509
Congrats to @FishAudio on the launch of Drama 3!
We’re excited to join as a day-one launch partner—and marking the occasion by bringing their full model lineup to the Eragon gateway.
Text-to-speech, speech-to-text, and the new Drama 3. All live today.
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment “DRAMA” within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
1
8
374
SF Tech Week doesn’t have to end with another conversation over a name badge.
Introducing Out of Office by Eragon × @hellyeah_ai. Think underground techno with immersive visuals.
Hapki, who’s opened for The Chainsmokers and Black Coffee, joins @Doudoujay of @hellyeah_ai and @contextconor of @hyperspell on the decks.
With friends from @neobrowser, @hyperspell, @astorie_ai, @SSG_Accelerator, @collov_ai and @OpusClip.
Oct 9th | RSVP in thread | See you on the dance floor.
SF Tech Week has enough rooms for networking.
On Oct 9, we are bringing the underground rave scene back to SF, introducing Out of Office-The Techno Sound, a one-of-a-kind multimedia immersive rave experience in the Bay Area.
Come rave with the cutest crowd in SF 🤫
* LinkedIn networking is strictly banned on the dance floor.
👇 RSVP link in thread
1
1
4
510
Today we are publishing our first collaborative research with @Eon_io_
We gave an Eragon agent a readable guide to a company's systems, then let GEPA rewrite that guide from the agent’s own failed traces.
GPT-5.6 Luna is the cheapest model we tested, and cold it was the weakest of the GPT and Claude models: 7/35.
Post GEPA: 29/35.
7
6
45
18,012
The fair objection: GEPA cost $8.54 to run. Does that eat the savings?
GEPA cost is charged once, then spreads across every question the package answers.
Cost per correct answer, optimization included: GPT-5.6 Luna passes GPT-6 Astra cold at 31 questions, which falls inside the measured range.
1
1
4
498
The company we tested on is Fairhaven Group, a synthetic enterprise generated by Era, with Salesforce, HubSpot, Zendesk, Jira, Slack and Gong on vendor-compatible MCP and REST interfaces.
Thank you to the team at @Eon_io_ for providing the ground of our research. Try out Era by Eon now at console.era.eon.io/ for free today!
1
331
