Pricing
Most of your work is cheap. Keeping it that way is the product.
Sending every prompt to a frontier model is the most expensive possible way to run AI, and almost none of your day needs one. The router's job is to keep everyday work in the cheap lanes and spend the frontier only on the moments that earn it. That is not a discount — it is where the money actually goes.
The spread · measured
The same work costs $0.71 or $43.59 per thousand tasks. Which one you get is a routing decision, not a plan you picked.
This section used to be arithmetic on the providers' published list prices, under a promise that it would be replaced by our own same-test measurement once the qualification sweep published. It has, so here it is: 52 priced models put through one task set of 11 domains under identical conditions, with cost computed from the tokens each one actually spent. The spread is not a lane chart we drew. It is what the library measured.
One tick per priced model, cost per task. Logarithmic, because the library spans 4 decades and a linear axis would draw 51 of these as hairlines beside the dearest one. 2 further models measured at zero — a free tier, not a missing figure — and cannot sit on a log axis: gemma-4-26b-a4b-it and gemma-4-31b-it, which scored 31.6% and 42.4%.
- Cheapest priced modelGroq · llama-3.1-8b-instant$0.043
64.0% pass over 114 scored tasks · $0.000043 per task · 10 of 11 domains. The floor of the measured library.
- Cheapest above 90% passGoogle · gemini-3.1-flash-lite$0.71
91.1% pass over 124 scored tasks · $0.000714 per task. The cheapest row in the sweep that cleared the threshold — and it cleared it on the full task set, not on the spine.
- Dearest model measuredAnthropic · claude-fable-5$43.59
72.7% pass over 11 scored tasks · $0.0436 per task. Eleven tasks is the core spine only. Read that pass rate with the depth, in both directions.
| Cost per task | Models | Median pass | Best pass | Tasks behind them |
|---|---|---|---|---|
| $0.00001 – $0.0001 | 1 | 64.0% | 64.0% | 114 |
| $0.0001 – $0.001 | 9 | 80.6% | 91.9% | 46–124 |
| $0.001 – $0.01 | 29 | 85.7% | 100.0% | 20–124 |
| $0.01 – $0.1 | 13 | 90.9% | 100.0% | 11–21 |
The decade a model's price falls in, and how the field in that decade scored. The bands are whole decades, not brackets chosen to make a point, and each one publishes the range of task counts behind it — a 90% median over 11 tasks is not a 90% median over 124.
$0.71 against $43.59 per thousand tasks — 61× — on work graded the same way. The cheapest model that cleared 90% did it over 124 scored tasks; the dearest model in the library ran 11 and passed 72.7% of them. End to end the priced library spans 1,014×. Compare only the 15 models that all ran the full suite (95–124 tasks each), where the means are taken over comparable work, and it is still 77× — with the dearest of those, qwen/qwen3.6-27b, holding the lowest pass rate in the entire sweep at 15.7%. Price does not predict the score in either direction. That is the routing problem, measured.
What this is, and is not. Measured 2026-08-17, suite qualification-v1, 54 models on one task set of 11 domains — not list prices. Costs are computed from measured token counts at published provider rates: arithmetic on a measurement, not an invoice, and they exclude negotiated and free-tier pricing. Coverage is not quite even either. 50 of 54 models returned a scoreable answer in every domain; the other 4 did not, and one of them is the cheapest row above — so their domain counts are printed on the rows rather than averaged away. Depth is uneven. Cost per task is a mean over that model's run, and the runs range from the 11-task core spine to the full set, so this is cost per task within one suite rather than a paired task-by-task comparison — the export publishes cost per model, not cost per task per model. One run is a snapshot.
The same spread, in Sparks · illustrative
What that costs you, in the unit you are billed in.
The sweep prices models in dollars per task. You are billed in Sparks per answer, and the sweep cannot produce that number — it measured models, not our billing. So this second chart stays, and it stays labelled for what it is: the lane map, read off the providers' public list prices. It is the bridge between the measurement above and the allowance below, not a second measurement.
- Fallback / speedGroq (Llama)0.03–0.1 sp
Greetings, lookups, one-line answers.
- EverydayGemini Flash0.1 sp
Ordinary reasoning and multi-modal input.
- WorkingDeepSeek0.3 sp
Real tasks — drafting, review, structured output.
- CodeKimi K2.7-Code0.4 sp
Generation and repair against a repo.
- Long contextQwen-Max0.6 sp
Documents that don’t fit anywhere else.
- FrontierClaude Opus3–6 sp
Reserved for the moments that earn it.
The frontier lane is about 45× the everyday lane. That is the number the axis above is hiding: the scale is logarithmic, because on a linear one the five cheaper lanes would be hairlines next to Opus and you could not read them at all. Illustrative, from the providers’ public list prices — not our measurements. The measured version of this spread leads the page above; this table is the lane map in the unit you are billed in.
What 1,000 Sparks buys · measured
Enough that you stop counting.
Both numbers above are measured. The divisor is 0.1148 sp — the mean of the 87 everyday answers that were charged anything at all, out of 120 logged between 2026-08-01 and 2026-08-18. The other 33 cost nothing, and they are left out on purpose: averaging free answers in drags the divisor to 0.083 sp and pushes the headline over 12,000, which would be a better number and a worse one. Someone spending a Premium allowance is not getting free-tier answers.
The other end of the same allowance: at the dearest answer we have ever logged — 0.8 sp, Regulated guidance, gpt-4o-mini — 1,000 Sparks is 1,250 answers. So the honest range for a month is roughly 1,250 to 8,708, depending entirely on what you ask. This section used to say it would stop being arithmetic once a measurement landed. It has: 145 real answers, 2026-08-01 to 2026-08-18. What it still is not is a month of YOUR usage, and no division can be that.
The plans
Start free. No card.
You can use Keenoble today without paying anything and without entering a card. Premium is what you move to when the free allowance stops being enough.
Premium
Mainstream Keenoble plan with 1,000 Sparks credits designed to cover normal monthly chat, preview, and code usage through smart model routing.
- 1,000 Sparks credits
- smart chat model routing
- preview mode
- code precision mode
Pro
Higher-context Keenoble plan designed for 5x usage, project orchestration, multi-project context, and a verified project pipeline.
- 5x Premium usage
- project orchestration
- full context over multiple projects
- high-context personal project agent
Agentic Workforce
Custom recurring skill-workforce orchestration for source-watch loops, eval loops, PR loops, verification gates, and human approval workflows.
- custom automated skill workforce
- recurring missions
- source-watch and eval loops
- PR loops with human approval gates
The full 335-model library (counted, not rounded) is included in every Keenoble plan — there is no tier that hides models behind it. Image, video and music generation are catalogued in the lab and will arrive as credit add-ons on top of Premium; they are coming, not claimed. Agentic Workforce is Keen OS Command — see /workforce. Keen Labs prices are founder-reported, verified 2026-05-21.
The math
Cancel the stack.
Replicating what one Keenoble seat gives you costs $105 a month across 5 logins whose tools have never met each other. Then add pay-as-you-go API keys for the models those subscriptions don’t include — DeepSeek, Kimi, Qwen, Grok — and live data feeds on top. You’re past $150/month and you still have zero shared memory, no routing, and five tabs of copy-paste.
Without Keen
$105/mo + API keys
With Keenoble · Premium
€49/mo · 1,000 Sparks
≈ 8,708 everyday answers at the 0.1148 sp we actually charged, or 1,250 at the dearest answer we have logged (0.8 sp). Measured, 2026-08-01 to 2026-08-18 — not a forecast of your month.
- Chat — smart-routed across the measured 335-model library
- Preview — idea to deployed, context inherited
- Code — repo-connected, shared memory
- Live tools — market data, web, weather, sources cited
- Receipts on every answer · cancel anytime
One currency, one receipt: you pay for work you actually received, and a run that fails its gate doesn’t bill. Keenoble is designed to consolidate AI tools for thinking, writing, coding and building. Image, video and music generation are already catalogued in the lab and will be released as credit add-ons on top of Premium — not claimed as shipped until they are. Competitor prices from public pricing pages, 2026-05-21.
Pilot · five seats
Or skip the plans and take a seat.
One month of Premium, free, with a direct line to the founder. Five seats, because five is how many direct lines one person can actually keep.