On your hardware
Free
Free local fine-tuning with the open-source package.
- Cost
- $0 with no account or card required
- Training
- LoRA tuning with frozen base and 4-bit exports on your hardware
Code · Vendor site 2026-10-04
Cactus Compute builds automation foundation models for tiny on-device targets such as phones, wearables, robots, and microcontrollers. Its Needle models run locally in small 8–29 MB footprints with tool-calling and structured extraction, plus hosted fine-tuning on Cactus GPUs.
Edge and embedded teams that need compact local models and optional hosted fine-tuning for Needle on device constraints.
Lowest paid monthly hosted tier is Pro at $99/month billed monthly on Needle fine-tuning pricing.
Cactus positions Needle as a compact automation foundation model family for on-device tool calls and extraction, with optional hosted fine-tuning tiers.
What it can do
Unknown is different from unavailable. Each fact carries its own evidence.
| Capability | Value | Evidence | Checked |
|---|---|---|---|
| Model size | Homepage cites 2-bit, 8–29 MB Needle models running locally up to about 4k tokens/sec decode. | Facts sourced | 2026-10-04 |
| Device targets | Marketing lists mobile, wearables, AR glasses, smart home, robots, cars, PCs, consoles, TVs, and microcontrollers. | Facts sourced | 2026-10-04 |
| Hosted fine-tuning | Needle pricing offers hosted Starter ($19 once) and Pro ($99/month) GPU fine-tuning plans. | Facts sourced | 2026-10-04 |
Understand the total cost
On your hardware
Free
Free local fine-tuning with the open-source package.
Pro
$99.00 /mo
$99/month billed monthly on hosted Pro fine-tuning.
Cosine, hosted at cosine.sh, is a sovereign AI lab training specialized coding agents and models for secure environments. It offers Lumen Scout, Outpost, and Sovereign models plus CLI and cloud surfaces with transparent credit-based plans for developers and regulated teams.
Explore toolIt supports local and cloud LLMs, codebase indexing on paid tiers, and a free Basic plan with your own API keys.
Explore toolAmika provides multiplayer cloud workstations where developers and coding agents share VMs loaded with their preferred agents. Customers bring their own LLM keys and pay for platform access plus sandbox compute credits with concurrent sandbox limits per tier.
Explore toolThe practical questions
Cactus Compute builds automation foundation models for tiny on-device targets such as phones, wearables, robots, and microcontrollers. Its Needle models run locally in small 8–29 MB footprints with tool-calling and structured extraction, plus hosted fine-tuning on Cactus GPUs.
Local fine-tuning with the open-source package is free at $0 with no account required.. This record lists ongoing free access; check the plan limits before starting.
The listed monthly price is $99.00 USD at monthly billing. See the plan cards for entitlements, billing commitments and seat minimums.
Not confirmed as a public HTTP API; offering centers on open-source CLI and hosted fine-tuning platform access.. API access and subscription access may have different terms; consult the linked sources.
Lowest paid monthly hosted tier is Pro at $99/month billed monthly on Needle fine-tuning pricing.
No retained pricing changes yet. A current price alone does not establish a historical trend.
Reviewed vendor source
Read original source ↗Facts apply to the named version and check date. Send a sourced correction if something changed.