Prediction Held up AI assessment confidence: 92% certainty 4/5 debate potential 2/5

Webster: Equivalent open-source reasoning models will arrive within weeks

Ian Webster · How to use DeepSeek safely · Feb 28, 2025 · at 13:42

Ian Webster, founder of AI security testing startup Promptfoo, advises enterprise leaders on the a16z Podcast regarding open-source alternatives to DeepSeek.

0:00 / 0:13exact quote · 13.2s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“What I've been telling most people who ask is like, let's just wait a few weeks and there will be an open source model that, that implements this reinforcement learning technique and you, you'll, you'll get great reasoning on par with what we see from deep seek.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

An open-source AI model implementing DeepSeek's reinforcement learning technique with reasoning capabilities on par with DeepSeek will be released within a few weeks.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

On March 5–6, 2025, well within the four-week window following the February 28, 2025 prediction, Alibaba released QwQ-32B under the permissive Apache 2.0 open-source license. QwQ-32B implemented a scaled reinforcement learning post-training methodology using outcome-based rewards and verifiers directly following the RL reasoning paradigm popularized by DeepSeek-R1. Alibaba's evaluation showed QwQ-32B achieving reasoning performance on par with or exceeding DeepSeek-R1 across major benchmarks including AIME, MATH, GPQA, and LiveCodeBench.

Alibaba released QwQ-32B on March 5–6, 2025, under an Apache 2.0 open-source license.
QwQ-32B used a multi-stage reinforcement learning pipeline with verifiable rewards, directly implementing the DeepSeek-R1 reasoning paradigm.
QwQ-32B demonstrated performance competitive with or surpassing DeepSeek-R1 on key reasoning benchmarks such as AIME, MATH, GPQA, and LiveCodeBench.

Checked against

[1] arxiv.org [2] time.com [3] medium.com [4] learnprompting.org [5] github.io [6] wikipedia.org [7] github.com [8] llmwatch.com

open source reasoning model released March 2025 DeepSeek R1 "QwQ-32B" "reinforcement learning" DeepSeek GRPO open source AI reasoning model DeepSeek RL GRPO March 2025 QwQ 32B release date Alibaba Open R1 Hugging Face release date 2025

assessed as of 2026-07-28

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

DeepSeek

⚑ Spot an error? Report it

Webster: Equivalent open-source reasoning models will arrive within weeks

Turn any episode into a week of clips.