🔍 Read the full analysis: What Makes BotPsy’s On-Device Personal Assistant Different? 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 over the base model on several benchmarks, but the results have not been independently validated in the supplied material, and practical device performance remains unclear.
Tether AI Research has introduced BotPsy, a 0.8-billion-parameter model based on Qwen3.5-0.8B that is designed to call tools and handle agent workflows, including through on-device deployment. The original analysis of BotPsy describes its capabilities. The company reports improved scores over the base model on several tool-use benchmarks, but those figures are company evaluations; the supplied material provides no independent testing or evidence of performance in everyday use.
BotPsy is positioned as a personal assistant for routine, low- to medium-difficulty tasks, rather than a general-purpose frontier model. The examples described by Tether AI Research span web research, travel and lifestyle requests, commerce and finance, business operations, device controls, documents, calendars and messaging. The model is distributed as weights, with the source listing safetensors and GGUF formats in bf16, Q8 and Q4.
In company-reported results, BotPsy scored 44.0 on BFCL v4, compared with 23.4 for Qwen3.5-0.8B-Think; 58.7 on ACEBench, compared with 30.1; and 51.5 on τ-bench, compared with 23.2. Tether says the τ-bench score is the mean of its retail, telecom and banking subsets. It also says BotPsy led models with up to 1.2 billion parameters across six tested tool-use benchmarks.
Against the larger Qwen3.5-2B-Think, the reported results were mixed: BotPsy scored higher on five of the six tests, including BFCL v4, ACEBench, τ-bench, ToolSandbox and BrowseComp-Plus. The larger model scored higher on Claw-Eval, at 48.9 versus BotPsy’s 46.5. The source also reports gains in instruction-following and some knowledge tests, while describing 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.
What Local Assistant Tasks Could Gain
BotPsy’s stated distinction is its combination of tool use and a relatively small model size. If it runs effectively on a user’s device, some routine requests could be handled without sending every prompt to a cloud service. That could give users more control over where certain data is processed and reduce reliance on external services for tasks a smaller model can manage.
Those are potential benefits, not established outcomes. The announcement supplies no independent cost comparison, and it does not show how BotPsy performs across actual phones or computers. Running locally does not, by itself, establish that an assistant is private, reliable or safe: those properties depend on the software around the model, the tools it can access and the way its outputs are checked. Tether itself says the model can make mistakes and is not intended for high-stakes use.
Tool selection matters because an assistant can do more than generate text: it may take actions through connected services. Tether reports an 86.1 score on BFCL v4 hallucination tests, which assess abstention when no suitable tool is available. The company also reports a 31% unsafe-response rate under prompt-injection attacks, down from 46% for the base model. These are bounded test results, not proof of safety across deployments or attack conditions.
Products Worth Considering
Apple Watch Series 11 GPS + Cellular, 46mm Jet Black Aluminum Case with Black Sport Band (M/L)
Apple iPad 11-inch (A16), 128GB, Wi-Fi, Silver
EcoFlow DELTA 3 Max Portable Power Station, 2048Wh LFP Battery
As an affiliate, we earn on qualifying purchases.
From Qwen3.5-0.8B to BotPsy
BotPsy is built on Qwen3.5-0.8B, with additional emphasis on function calling and agent workflows. Tether’s case for a specialized compact model is that it may perform routine assistant work without requiring a larger cloud model. The proposed trade-off is between the capabilities and resources of a larger system and the possibility of running selected tasks locally.
The announcement’s benchmark comparisons suggest the model has been adapted for tool-related evaluations, but they do not settle how it will behave in a complete assistant. Real use can involve ambiguous requests, changing tool interfaces, multiple steps and interruptions. The supplied source includes selected scores and comparisons but does not provide an independent evaluation, full testing protocols or user deployment results.
“BotPsy is a “personal assistant” that uses tools to perform a variety of simple tasks for the user.”
— Tether AI Research
Real-Device Results Remain Unknown
The supplied material does not show independent reproduction of the benchmark scores or explain the full testing protocols. It also does not establish whether the results will hold across different implementations, tools or user requests. Benchmarks can measure defined tasks, but they do not by themselves show that an assistant will reliably complete longer, less predictable workflows.
Practical deployment details are also missing. The announcement does not specify BotPsy’s memory use, speed or power consumption on representative devices, or compare end-user costs with cloud models. It gives no quantified account of the reported weaknesses as instructions accumulate or in lower-resource languages. The source also does not identify a release date, update schedule or support commitments, so availability and maintenance expectations are unclear.
Evidence Needed From Deployments
The next useful evidence would be independent benchmark reproduction and reports from deployments on representative phones, computers or other devices. Those assessments could clarify latency, memory and energy requirements, along with how consistently BotPsy calls tools, recovers from failures and handles longer sequences of instructions.
Potential users will also need documentation on supported runtimes, tool integrations, safeguards and model updates. Tether’s announcement, as presented in the supplied material, does not give a timeline for further evaluations or releases. Until those details and practical results are available, the reported scores offer an initial indication of performance on selected tests, not a demonstrated measure of reliability in everyday use.
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.
What makes BotPsy different from a typical cloud assistant?
Its design targets tool-based tasks in a compact model that the company says can be deployed on-device. Whether it reduces cloud use in a particular product depends on the deployment and supported tools.
How well did BotPsy perform in tests?
Tether reports improved scores over its base model on several benchmarks, including BFCL v4, ACEBench and τ-bench. The results are company-reported; the supplied material does not include independent validation.
Is BotPsy proven safe for sensitive tasks?
No. Tether says the model can make mistakes and is not intended for high-stakes use. Its reported safety-related test figures do not establish safety in every setting.
When can people use BotPsy on their devices?
The supplied information does not state a release date or provide details on supported devices, practical hardware requirements or update commitments. Availability remains unclear.
Primary source: Hugging Face · via ThorstenMeyerAI.com
