WizardLM
Evol-Instruct LLM family — ex-Microsoft team, now dormant open weights
Overview
WizardLM is an open instruction-tuned LLM family (WizardLM, WizardCoder, WizardMath) built on Evol-Instruct, which evolves training instructions into progressively harder variants. It has been effectively dormant since early 2024: the Microsoft-based team departed for Tencent, where it now builds Hunyuan models, and the original repository remains available but unmaintained.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
WizardLM popularised Evol-Instruct — automatically evolving training instructions into harder variants — and its open models punched above their weight in 2023–24. The family (WizardLM, WizardCoder, WizardMath) remains free open weights, useful for research, fine-tuning experiments and on-prem deployments where data can't leave the building. But the story has moved on: Microsoft pulled WizardLM-2 hours after its April 2024 release over skipped toxicity testing, and in May 2025 the team decamped to Tencent's Hunyuan lab, ending development under the WizardLM name. There's no hosted service or API — you run the weights yourself, which takes GPUs and ML skills. Today it's a well-documented reference lineage for instruction-tuning research rather than a competitive chat or coding model; general users wanting a ready-made assistant should look elsewhere.
What people say
In April 2024, Microsoft published WizardLM-2 and pulled it within hours. The team had skipped a required toxicity-testing step before release, and the takedown became the strangest chapter in the family's history — mirrors of the Apache 2.0 weights survived on Hugging Face and OpenRouter anyway. A stranger chapter followed in May 2025, when TechCrunch reported that the Beijing-based WizardLM group, led by researcher Can Xu, had left Microsoft entirely for Tencent's Hunyuan lab. Their first model there shipped under the Hunyuan name, not WizardLM.
The underlying research still holds up. Evol-Instruct, the method that made the family's name, uses an LLM to rewrite training instructions into progressively harder variants — more constraints, deeper reasoning chains, combined tasks — so a model sees a far wider difficulty range than human-written data provides. WizardLM-30B hit 97.8% of ChatGPT's score on the project's own skill evaluations, WizardCoder-33B posted 79.9 pass@1 on HumanEval, and WizardMath-7B reached 83.2 on GSM8K — striking numbers for open models in 2023–24, though all self-reported.
What's left in 2026 is weights and a paper trail. The GitHub repo is effectively frozen, no hosted API ever existed, and the open-model field — Qwen, DeepSeek, Llama — has moved several generations past. If you want to study how instruction difficulty affects fine-tuning, the Evol-Instruct code and datasets are still worth pulling. If you want a chatbot or coding model to deploy this year, this isn't the shortlist. Treat WizardLM the way you'd treat a well-cited paper: read it, learn from it, run something newer.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- WizardLM's Evol-Instruct creates instructions that are 4x more complex on average than the original training data, teaching models to handle requests that simple instruction datasets never include.
- The WizardLM family (WizardLM, WizardCoder, WizardMath) was produced by a team of 3 Microsoft Research interns in under 6 months.
- WizardLM-70B was the first open model to achieve GPT-3.5-level performance on the MT-Bench conversational evaluation.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including WizardLM's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with WizardLM unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to WizardLM directly →
Spotted something out of date? Suggest an update →
Alternatives to WizardLM
More in LLMs & Chat
