🔍 Read the full analysis: How BotPsy Runs A Personal Assistant On A Sub-Billion-Parameter Language Model on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Tether AI Research has introduced BotPsy, a 0.8-billion-parameter model based on Qwen3.5-0.8B for tool use and agent workflows, with on-device deployment as a stated use case. The company reports gains on six benchmarks, but the figures have not been independently verified and do not show how BotPsy performs in everyday deployments.
Tether AI Research has introduced BotPsy, a 0.8-billion-parameter language model based on Qwen3.5-0.8B and designed for function calling and agent workflows, as described in the original analysis. The company says it is aimed at low- to medium-difficulty personal assistant tasks and can be deployed on-device, but its published benchmark results are company-reported and do not establish how reliably it will work in everyday use.
BotPsy is intended to use tools for tasks such as web research, calendars, messaging, document handling, device controls, and travel, commerce, finance, and business requests. Tether AI Research describes the model as a personal assistant for relatively simple work rather than a general-purpose system for demanding or high-stakes decisions. The company also cautions that the model can make mistakes and is not intended for high-stakes use.
In results attributed to Tether AI Research, BotPsy scored 44.0 on BFCL v4, compared with 23.4 for its Qwen3.5-0.8B-Think base model. It scored 58.7 on ACEBench versus 30.1 for the base, and 51.5 on τ-bench versus 23.2. Tether says the τ-bench result is the mean of its retail, telecom, and banking subsets. The company reports that BotPsy led models with up to 1.2 billion parameters across six tested tool-use benchmarks.
Comparisons with the larger Qwen3.5-2B-Think were mixed: BotPsy scored higher on five of the six cited tests, including BFCL v4, ACEBench, τ-bench, ToolSandbox, and BrowseComp-Plus. Qwen3.5-2B-Think scored 48.9 on Claw-Eval, against BotPsy’s 46.5. Tether also reports an 86.1 score on BFCL v4 hallucination tests and a 31% unsafe-response rate under prompt-injection attacks, down from 46% for the base model. These figures describe the company’s testing, not safety across all uses.
BotPsy brings tool use to a compact model
Tether AI Research says its small assistant is designed for routine agent workflows and on-device use. Its benchmark results are developer-reported and do not establish everyday reliability.
Built for routine assistant work
The model is tuned for function calling and agent workflows, rather than frontier-scale general performance.
Find and gather information
Example uses include web research and travel requests.
Handle everyday workflows
It is presented for commerce, business operations, documents, calendars, and messaging.
Run on-device
Local deployment could reduce reliance on cloud services for routine tasks.
Reported results improved over its base
Tether AI Research reports higher scores for BotPsy on three named benchmarks.
| Benchmark | BotPsy | Qwen3.5-0.8B-Think |
|---|---|---|
| BFCL v4 | 44.0 | 23.4 |
| ACEBench | 58.7 | 30.1 |
| τ-bench | 51.5 | 23.2 |
The larger-model comparison was mixed
BotPsy scored higher on five of six tool-use tests against Qwen3.5-2B-Think.
- BFCL v4
- ACEBench
- τ-bench
- ToolSandbox
- BrowseComp-Plus
- Qwen3.5-2B-Think scored 48.9 on Claw-Eval, compared with BotPsy’s 46.5.
Limits matter for deployment
The supplied material describes areas where performance may fall short and leaves practical questions open.
- Results come from the developer; independent reproduction is not provided.
- Performance drops as instructions accumulate and lags in lower-resource languages.
- Hardware needs, memory use, power use, and real-world reliability are not established.
- Tether says BotPsy can make mistakes and is not intended for high-stakes use.
What Local Assistant Tasks Could Gain
A model with 0.8 billion parameters could be easier to run on some personal devices than larger assistant models, if its software and hardware requirements permit. Local execution may let users handle some routine requests without sending every prompt to a cloud service. That could matter for privacy, availability, and operating costs, although the announcement does not provide verified cost comparisons or show that users can keep all data on-device in a complete workflow.
The results also speak to a technical challenge beyond generating text: an assistant must select the right tool, pass it suitable instructions, and respond appropriately when no available tool can do the job. Tether’s benchmark results suggest improvement over the base model on its selected tests. They do not establish that BotPsy can consistently complete longer, unpredictable tasks or safely manage sensitive information. Its potential value will depend on performance outside the reported evaluations.
Products Worth Considering
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
DJI Mini 4 Pro Fly More Combo Drone with DJI RC 2
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
As an affiliate, we earn on qualifying purchases.
BotPsy’s Qwen3.5 Foundation
BotPsy is built on Qwen3.5-0.8B and is presented as a specialized model for function calling and agentic workflows, rather than as a frontier-scale general assistant. The source says its weights are distributed in safetensors and GGUF formats, with bf16, Q8, and Q4 options. The announcement frames smaller models as a way to handle routine tasks with less dependence on cloud services; that is the developer’s rationale, not a measured comparison of total costs or privacy outcomes.
The source also reports gains in instruction-following and some knowledge benchmarks, while flagging weaker performance as instructions accumulate and in lower-resource languages. It does not quantify those limitations in the supplied details. No independent evaluation or user deployment results are provided, so the reported scores should be read as the developer’s test results rather than a settled measure of real-world performance.
“BotPsy is a “personal assistant” that uses tools to perform a variety of simple tasks for the user.”
— Tether AI Research
Independent Tests and Device Needs
The supplied material does not identify an independent evaluation, publish full testing protocols, or show results from people using BotPsy in deployed assistants. It is not clear whether the benchmark scores will reproduce across different runtimes and implementations, or how closely they predict reliability with changing tools and varied user requests.
Practical requirements are also unspecified: the announcement gives no measured details on memory use, power consumption, latency, or supported hardware. The extent to which a complete assistant workflow can run locally, including its tools and any related services, is not established. Tether reports weaker performance as instructions accumulate and in lower-resource languages but does not quantify those gaps in the supplied account. Release timing, ongoing support, and update commitments are also not stated.
Evidence Needed for Deployment
The next useful evidence would be independent reproduction of the benchmark results and tests on representative phones, computers, and other devices. Such evaluations could measure response time, memory and energy demands, tool-call reliability, and performance across longer workflows, including failures and recovery.
People considering the model for practical tasks will also need documentation on supported runtimes, safeguards, and update practices. Tether AI Research has not provided a timeline for further evaluations or releases in the supplied material. Until those details and deployment results are available, the reported benchmark gains remain an early indication of capability, not proof of dependable everyday use.
Key Questions
What is BotPsy?
BotPsy is a 0.8-billion-parameter model based on Qwen3.5-0.8B. Tether AI Research designed it for function calling and agent workflows, including personal assistant tasks.
Can BotPsy run on a personal device?
On-device deployment is a stated use case, and the source lists model weights in safetensors and GGUF formats. The announcement does not provide device-specific requirements or performance measurements, so the range of devices it can run on is unclear.
How strong are BotPsy’s reported results?
Tether AI Research reports higher scores than its Qwen3.5-0.8B-Think base model on several tool-use benchmarks. The results are company-reported; the supplied material does not provide independent verification or establish performance in everyday use.
Is BotPsy suitable for high-stakes decisions?
No such use is recommended in the source material. Tether AI Research says the model is intended for low- to medium-difficulty tasks, cautions that it can make mistakes, and says it is not intended for high-stakes use.
Primary source: Hugging Face · via ThorstenMeyerAI.com
![AI eats the world (Spring 26) [pdf]](https://aismasher.com/wp-content/uploads/2026/05/ai-eats-the-world-spring-26-pdf-featured-560x300.jpg "AI eats the world (Spring 26) [pdf] 12")
