The first Agnes AI text model on the independent Artificial Analysis leaderboard runs the average task for about three cents, a fraction of what comparable models cost, without giving up the capability to do real work.
— Agnes AI today debuted Agnes 2.5 Pro Alpha on the independent Artificial Analysis leaderboard — its first text model to be measured there, joining the company’s image and video models already ranked in the top ten. The Singapore-based lab, which trains its own models in-house, posted a result that reframes a tradeoff every builder of agentic products knows well: the models capable enough to run a task end to end are the ones that make the monthly bill flinch-worthy. Agnes 2.5 Pro Alpha was built to end that choice — pairing frontier-grade capability with a cost that doesn’t force teams to ration it.
Starting near the top, not the bottom
On the Artificial Analysis Intelligence Index v4.1, which blends nine separate evaluations, the model scored 39, the standout in its published cohort by a wide margin. The nearest competitors, Cohere’s Command A+ at 23 and Mistral’s Devstral 2 at 19, land well behind. For an alpha, that is an unusual place to start.
But can it actually do the job?
A cohort-topping average is not why a developer would switch. What matters is whether the model can do the actual work of a coding agent, and the results say it can. It writes code well, holding its own on the Coding Index (58.8) against pricier competition like GPT-5.4 Plus and DeepSeek V4 Flash, and just behind Nex-N2-Pro. It doesn’t just produce code that looks right, it operates the machine, scoring 67% on Terminal-Bench v2.1 for running commands and working a task end to end, ahead of DeepSeek V4 Pro. And it handles the hard reasoning, reaching 88% on GPQA Diamond, a set of graduate-level science questions built to resist search, close to the leading score of 94%.
Write the code, drive the machine, think through the problem. That’s the whole job, and it’s where the model is built to perform.
The number that decides things
Capability is half the story. The other half is what it costs to put that capability to work every day. On GDPval-AA v2, Artificial Analysis’s evaluation of real-world tasks, Agnes 2.5 Pro Alpha completed the average task for about three and a half cents. Put that on a monthly bill: ten thousand tasks a month is $342 on Agnes 2.5 Pro Alpha. The same ten thousand tasks cost $12,300 on GPT-5.6 Sol and $37,000 on Claude Fable 5.*
Two things drive that gap. Agnes 2.5 Pro Alpha is priced low per token, at $0.45 per million input tokens and $0.90 per million output. And it does not overthink: on the same evaluation it reached its answers in 26 turns and 38,000 output tokens, where some models burned three to four times as many. For a team running agents continuously, that difference compounds every single day.
“Most of the world’s developers are priced out of the best AI, not because they lack the skill to use it but because the meter never stops running,” said Bruce Yang, Founder of Agnes AI. “We built Agnes to change that arithmetic. Frontier capability should not have to cost like the frontier. This is the earliest version of that idea, and we are putting it in public so anyone can hold us to it.”
This is an alpha, and we mean it
Agnes 2.5 Pro Alpha is exactly what its name says: an alpha, the earliest version of the model to reach Artificial Analysis, and not the finished product. It will be updated over the coming months, and every update will be measured the same way, independently and in public, against everyone else. Watch the leaderboard.
That public, measurable improvement is the point. Agnes AI’s mission is AI parity: making frontier-grade capability reachable for the individual developers, startups, and small teams who have been priced out of it. Free API access is where that starts.
Try Agnes on Agnes Code: https://agnes-ai.com/agnescode · API access: https://platform.agnes-ai.com/
* Cost figures are based on Artificial Analysis’s published per-task output-token counts at each provider’s list output price. Input token costs are not included.
About Agnes AI
Agnes AI builds full-modality foundation models, trained in-house across text, image, and video generation. Its mission is AI parity: making frontier-grade AI capability accessible at a cost that does not exclude individual developers, startups, and small teams.
Discord:https://discord.com/invite/AJEfcvvE3N
Contact Info:
Name: Victoria, Marketing Agnes AI
Email: Send Email
Organization: Agnes AI
Website: https://agnes-ai.com/
Disclaimer:
This press release is for informational purposes only. Information verification has been done to the best of our ability. Still, due to the speculative nature of the blockchain (cryptocurrency, NFT, mining, etc.) sector as a whole, complete accuracy cannot always be guaranteed.
You are advised to conduct your own research and exercise caution. Investments in these fields are inherently risky and should be approached with due diligence.
Release ID: 89199460
If you detect any issues, problems, or errors in this press release content, kindly contact error@releasecontact.com to notify us (it is important to note that this email is the authorized channel for such matters, sending multiple emails to multiple addresses does not necessarily help expedite your request). We will respond and rectify the situation in the next 8 hours.