🔍 Read the full analysis: What You Should Know About BotPsy’s On-Device AI Assistant on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Tether AI Research has introduced BotPsy, a 0.8-billion-parameter model based on Qwen3.5-0.8B and designed for tool use and agent workflows, including on-device deployment. The company reports gains on six tool-use benchmarks, but independent evaluations, real-device performance data and details about resource requirements are not provided in the source material.
Tether AI Research has introduced BotPsy, a 0.8-billion-parameter model built on Qwen3.5-0.8B for function calling and agent workflows, as described in the original analysis. The company says it can run on-device and handle routine tasks such as research, calendar work and device controls, but the available results are company-reported benchmarks, not independent evidence of how reliably it will perform in everyday use.
BotPsy is presented as a personal assistant that uses tools to carry out requests. The source describes possible uses across web research, travel and lifestyle requests, commerce and finance, business operations, document handling, calendars, messaging and device controls. Tether AI Research says the model is aimed at low- to medium-difficulty tasks, rather than high-stakes work or frontier-scale general-purpose performance, and acknowledges that it can make mistakes. Related context is available in this overview of AI tools and automation.
The company reports BotPsy scored 44.0 on BFCL v4, compared with 23.4 for its Qwen3.5-0.8B-Think base model. It also reports 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 the retail, telecom and banking subsets. These figures describe the company’s tests; the supplied material does not include an independent reproduction or full testing protocols. For further AI memory context, see this discussion of AI memory budgets.
Tether says BotPsy led models up to 1.2 billion parameters across six tested tool-use benchmarks. Results against the larger Qwen3.5-2B-Think were mixed: BotPsy scored higher on five named tests—BFCL v4, ACEBench, τ-bench, ToolSandbox and BrowseComp-Plus—while Qwen3.5-2B-Think scored higher on Claw-Eval, 48.9 to BotPsy’s 46.5. The source also reports improvement in instruction-following and some knowledge benchmarks, alongside weaker performance as instructions accumulate and in lower-resource languages.
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.
Local Assistants Need Real-Device Tests
If BotPsy performs reliably on supported hardware, its small size could let some assistant tasks run directly on a user’s device, potentially reducing how often requests and related data need to go to cloud services. That could matter to people seeking local control over routine work, and to developers weighing cloud access, operating costs and service availability. These are potential advantages, not established results: the source offers no measured cost comparison or user deployment data.
Tool-using models also have to choose actions and respond appropriately when tools cannot satisfy a request. Tether reports an 86.1 score on BFCL v4 hallucination tests, which it says assess abstention when no suitable tool is available. It also reports a 31% unsafe-response rate under prompt-injection attacks, down from 46% for the base model. Those figures are not proof of safety across real settings, tools or attack types. The company itself cautions against high-stakes use.
Products Worth Considering
LG UltraGear 32GX850A-B 32-inch 4K UHD OLED Gaming Monitor
Samsung Galaxy Tab S7 11-inch Android Tablet 128GB Wi‑Fi with S Pen Mystic Black
DJI Mini 3 Drone with DJI RC and 3-Axis Mechanical Gimbal
As an affiliate, we earn on qualifying purchases.
A Compact Model Built for Tool Use
BotPsy is based on Qwen3.5-0.8B and is distributed as model weights, with the source listing safetensors and GGUF formats in bf16, Q8 and Q4. Its stated focus is function calling and agent workflows: the model is meant to select and use tools for tasks, rather than simply generate text without taking actions.
Tether’s rationale is that a specialized, smaller model may handle some routine assistant work locally, without relying on a larger cloud model for every request. The source material gives benchmark tables and selected scores, but no independent evaluation, field trial results or detailed hardware testing. That distinction matters: a benchmark score does not establish how well a model will handle changing tools, ambiguous requests or longer sequences of actions.
“BotPsy is a “personal assistant” that uses tools to perform a variety of simple tasks for the user.”
— Tether AI Research
Performance and Safety Remain Unverified
The supplied information does not identify an independent evaluation or provide full test protocols, so it is unclear whether the benchmark results will reproduce across implementations. Nor do the scores establish reliability in real assistant workflows, where tools may change and requests can involve sensitive information.
Details are also missing on hardware requirements, memory use, power consumption and latency on phones or computers. The source notes weaker performance as instructions accumulate and in lower-resource languages, but does not quantify those limitations. It does not give a release date, specify support and update commitments, or provide a verified comparison of end-user costs against cloud-based assistants.
Deployment Data Is the Next Test
The most useful next evidence would be independent replication of the benchmark results and deployments on representative devices. Such testing could report latency, memory and energy use, tool-call reliability, and performance across longer workflows—not only isolated benchmark tasks.
People considering the model for practical use will also need documentation on supported runtimes, safeguards and update practices. Tether’s announcement does not set a timeline for further evaluations or releases, so availability and deployment details remain open.
Key Questions
What is BotPsy?
BotPsy is a 0.8-billion-parameter model from Tether AI Research, based on Qwen3.5-0.8B and designed for function calling and agent workflows. The company describes it as a tool-using assistant for routine tasks.
Can BotPsy run on a phone or computer?
Tether says the model can be deployed on-device, and the source lists model weights in safetensors and GGUF formats. It does not provide specific supported-device lists, hardware requirements, or measured performance on phones or computers.
Are BotPsy’s benchmark results independently verified?
The results in the supplied material are reported by Tether AI Research. It does not identify an independent evaluation or provide enough testing detail to establish whether the results will reproduce elsewhere.
Is BotPsy suitable for high-stakes tasks?
No. Tether says BotPsy is intended for low- to medium-difficulty tasks, cautions that it can make mistakes and says it is not intended for high-stakes use.
What is still unknown about BotPsy?
The source does not specify a release date, device-level memory or power needs, real-world reliability, end-user costs, or support and update commitments. It also does not provide independent safety testing or deployment results.
Primary source: Hugging Face · via ThorstenMeyerAI.com
