LangWatch used models served through Nebius AI Studio to compare banking support agents in simulation tests. Its repository lists DeepSeek V3.2, GLM-4.7, MiniMax M2.1 and GPT-oss 120B as Nebius-served models. The project combines Agno for the agent framework, Scenario for simulation testing and LangWatch for observability.
The comparison covers eight banking scenarios, including fraud investigation, escalation, lost cards and overdraft disputes. LangWatch reports running five rounds per scenario for each model and publishes evaluation results and latency measurements. Each model variant uses the same agent architecture, with specialized agents for conversation summaries, customer data exploration and response suggestions. Results are tracked in LangWatch for comparison across models.
LangWatch x Nebius: Comparing LLM Models with Agent Simulations
19 Oct 2026 · GitHub · Web page
LangWatch's repository describes agent simulation evaluations using Nebius AI Studio models. It reports completed runs across eight banking scenarios and publishes comparative results.
LangWatch x Nebius: Comparing LLM Models with Agent Simulations19 Oct 2026 · GitHub · Web pageLangWatch's repository describes agent simulation evaluations using Nebius AI Studio models. It reports completed runs across eight banking scenarios and publishes comparative results.
Nebius Builder Hour featuring LangWatch
19 Oct 2026 · LinkedIn · Web page
Nebius announces a partner discussion with LangWatch about agent testing during a Token Factory Builder Hour.
Nebius Builder Hour featuring LangWatch19 Oct 2026 · LinkedIn · Web pageNebius announces a partner discussion with LangWatch about agent testing during a Token Factory Builder Hour.
LangWatch provides an open-source platform for engineering and AI platform teams to test, evaluate and monitor AI agents and LLM applications. Operated by Reasoning Engine B.V., it combines simulated user conversations, automated quality evaluations, production tracing, prompt management and an AI gateway. Teams use it to debug agent behavior, compare models and prompts, track costs and latency, and catch regressions before deployment. The platform supports text and voice agents and is available as a hosted service or a self-hosted installation.
LangWatch launches Instant Evals for production history
Instant Evals evaluates stored conversations, traces or LLM calls against plain-language questions and returns matches with probabilities. Results can be exported for testing or fine-tuning, with up to 100,000 rows per run on paid plans.
The Go SDK reaches 1.0 with native instrumentation for providers including Anthropic, Gemini and Ollama. Its typed REST client covers prompts, datasets, traces, annotations and evaluations.
LangWatch adds connected-agent testing and local Langy execution
Python and TypeScript agents can become test targets without exposing a public URL. Langy can also work in an authorized local folder using the developer's existing tools.
LangWatch provides AI agent testing, evaluation and observability tools for engineering teams. · Developer tools
How they use Nebius Token Factory
LangWatch used models served through Nebius AI Studio to compare banking support agents in simulation tests. Its repository lists DeepSeek V3.2, GLM-4.7, MiniMax M2.1 and GPT-oss 120B as Nebius-served models. The project combines Agno for the agent framework, Scenario for simulation testing and LangWatch for observability.
The comparison covers eight banking scenarios, including fraud investigation, escalation, lost cards and overdraft disputes. LangWatch reports running five rounds per scenario for each model and publishes evaluation results and latency measurements. Each model variant uses the same agent architecture, with specialized agents for conversation summaries, customer data exploration and response suggestions. Results are tracked in LangWatch for comparison across models.
LangWatch x Nebius: Comparing LLM Models with Agent Simulations
19 Oct 2026 · GitHub · Web page
LangWatch's repository describes agent simulation evaluations using Nebius AI Studio models. It reports completed runs across eight banking scenarios and publishes comparative results.
LangWatch x Nebius: Comparing LLM Models with Agent Simulations19 Oct 2026 · GitHub · Web pageLangWatch's repository describes agent simulation evaluations using Nebius AI Studio models. It reports completed runs across eight banking scenarios and publishes comparative results.
Nebius Builder Hour featuring LangWatch
19 Oct 2026 · LinkedIn · Web page
Nebius announces a partner discussion with LangWatch about agent testing during a Token Factory Builder Hour.
Nebius Builder Hour featuring LangWatch19 Oct 2026 · LinkedIn · Web pageNebius announces a partner discussion with LangWatch about agent testing during a Token Factory Builder Hour.
LangWatch provides an open-source platform for engineering and AI platform teams to test, evaluate and monitor AI agents and LLM applications. Operated by Reasoning Engine B.V., it combines simulated user conversations, automated quality evaluations, production tracing, prompt management and an AI gateway. Teams use it to debug agent behavior, compare models and prompts, track costs and latency, and catch regressions before deployment. The platform supports text and voice agents and is available as a hosted service or a self-hosted installation.
LangWatch launches Instant Evals for production history
Instant Evals evaluates stored conversations, traces or LLM calls against plain-language questions and returns matches with probabilities. Results can be exported for testing or fine-tuning, with up to 100,000 rows per run on paid plans.
The Go SDK reaches 1.0 with native instrumentation for providers including Anthropic, Gemini and Ollama. Its typed REST client covers prompts, datasets, traces, annotations and evaluations.
LangWatch adds connected-agent testing and local Langy execution
Python and TypeScript agents can become test targets without exposing a public URL. Langy can also work in an authorized local folder using the developer's existing tools.
