Pinned Tweet
Same model. Better harness. +5 points on Terminal-Bench 2.1.
Opus 4.8 in Claude Code: 78.9%
Opus 4.8 in Nova Agent OS: 83.9%
Request access: novaagent.dev
We made Opus 4.8 perform like Fable 5.
Same model. No fine-tuning.
Terminal-Bench 2.1:
Opus 4.8 in Claude Code: 78.9%
Opus 4.8 in Nova Agent OS: 83.9%
Fable 5: 83.8%
Built together with our friends at @modal. The frontier isn't only the model anymore.
novaagentos.ai/news/how-nova…
Our own usage, Oct 21–Nov 19: $810.26 with Nova Agents vs an estimated $2,762.16 if everything ran on Claude Opus 5.5. About 70% lower.
1
2
We built Nova Agents to run our own company. Our agents keep their context across Claude Code, Codex and Cursor, and keep going when the laptop is closed.
It has worked well for us. If you want to try it, DM us and we'll send you 30 days free. No credit card needed.
2
6
New in Nova Agents: - Switch between Claude Code, Codex and Cursor in the middle of a session. Out of quota in one? Pick another and keep going.- Teams can turn on one shared brain for every member. Never tied to one AI. Watch the demo with CEO Carl: novaagents.com
2
7
Hit a usage limit mid-task? Keep going.
Nova Agents gives Claude Code, Codex and Cursor one shared brain. When your main model taps out, an open-weight model picks up the same task.
Never tied to one AI.
Watch the demo with CEO Carl: novaagents.com
1
33
Save your best model for the hard part.
Nova Agent OS picks a model for each task, shares context across agents and has reviewer agents check the work.
Smarter agents. Half the tokens. Never tied to one AI.
novaagent.dev
2
65
Switch tools. Keep the thread.
Claude Code, Codex and Cursor share one Brain in Nova Agent OS. Your projects, decisions and context stay with you when you switch.
Smarter agents. Half the tokens. Never tied to one AI.
novaagent.dev
3
9
Close the laptop. Your agents keep working.
Nova Agent OS runs Claude Code, Codex and Cursor in the cloud, 24/7. Reviewer agents check the work, and at a usage limit, an open model picks up the same session.
Tools come and go. Your agents shouldn't stop.
novaagent.dev
2
46
Same model. Next-generation results.
On Terminal-Bench 2.1, Claude Opus 4.8 scores 79% on its own and 84% running on Nova. That matches the next-generation frontier model.
No fine-tuning. A shared Brain, reviewer agents and smarter model routing.
novaagent.dev
1
18
Nova got Opus 4.8 to edge past Fable 5 on Terminal-Bench 2.1.
Modal Sandboxes let the team decouple Nova’s agent loop and automatic review from tool execution, with fast startup and changes deployed in seconds to keep iteration quick.
The result: 83.9%, up five percentage points from Opus 4.8’s Claude Code baseline and just ahead of Fable 5’s 83.8%.
We made Opus 4.8 perform like Fable 5.
Same model. No fine-tuning.
Terminal-Bench 2.1:
Opus 4.8 in Claude Code: 78.9%
Opus 4.8 in Nova Agent OS: 83.9%
Fable 5: 83.8%
Built together with our friends at @modal. The frontier isn't only the model anymore.
novaagentos.ai/news/how-nova…
Nova Agents retweeted
We made Opus 4.8 perform like Fable 5.
Same model. No fine-tuning.
Terminal-Bench 2.1:
Opus 4.8 in Claude Code: 78.9%
Opus 4.8 in Nova Agent OS: 83.9%
Fable 5: 83.8%
Built together with our friends at @modal. The frontier isn't only the model anymore.
novaagentos.ai/news/how-nova…











