Public story · 2026-03-15 · source-backed

Qwen3.5-9B Beats GPT-OSS-120B (13x Larger) on Consumer Hardware.

Issue

2026-03-15

Confidence

source-backed

Sources

Redaction

passed

Story

Alibaba's 9B model outperforms OpenAI's 120B on GPQA Diamond (81.7 vs 71.5), MMLU-Pro (82.5 vs 80.8), and multilingual MMMLU. Uses hybrid Gated Delta Network + sparse MoE with 262K native context. Apache 2.0 on HuggingFace. VentureBeat

Anthropic partners with OpenAI / Shared entities / What happened next
Moonshot cuts off new Kimi K3 signups after 48 hours
Linked by a graph relationship (Anthropic partners with OpenAI); both cover Alibaba, GPQA Diamond, MoE, OpenAI; picks up the Alibaba thread on 2026-07-20.
Hugging Face criticizes OpenAI / Shared entities / Same source domain / What happened next
Kimi K2.7-Code: a 1T-param open coding model with benchmarks nobody can check
Linked by a graph relationship (Hugging Face criticizes OpenAI); both cover GPQA Diamond, MMLU, MoE, VentureBeat; reported by the same outlet (venturebeat.com).
OpenAI partners with Google / Shared entities / Same source domain / What happened next
Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.
Linked by a graph relationship (OpenAI partners with Google); both cover Alibaba, MoE, OpenAI, VentureBeat; reported by the same outlet (venturebeat.com).
NVIDIA invested in OpenAI / Shared entities / Shared topic / What happened next
DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget
Linked by a graph relationship (NVIDIA invested in OpenAI); both cover Alibaba, Apache, MoE, OpenAI; overlapping topics (apache, context).
NVIDIA invested in OpenAI / Shared entities / What happened next
Meta ships Muse Glimmer, a 30B open agent model
Linked by a graph relationship (NVIDIA invested in OpenAI); both cover Apache, MoE, OpenAI, Qwen3; picks up the Apache thread on 2026-08-10.
LLM uses OpenAI / Shared entities / What happened next
Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.
Linked by a graph relationship (LLM uses OpenAI); both cover Apache, HuggingFace, MoE, Qwen3; picks up the Apache thread on 2026-04-02.
Hugging Face criticizes OpenAI / Shared entities / Shared topic / What happened next
Qwen3.8-27B is architecturally byte-identical to Qwen3.6-27B. Every gain came from training.
Linked by a graph relationship (Hugging Face criticizes OpenAI); both cover Alibaba, Apache, Qwen3; overlapping topics (apache, context).
NVIDIA invested in OpenAI / Shared entities / Shared topic / What happened next / Tension
Nemotron Cascade 2: NVIDIA's 30B-A3B MoE Outperforms Qwen3.5-35B
Linked by a graph relationship (NVIDIA invested in OpenAI); both cover HuggingFace, Qwen3; overlapping topics (beat, huggingface, larger).

Source trail

VentureBeat

Entities

Alibaba Apache Beats GPT Consumer Hardware Gated Delta Network GPQA Diamond HuggingFace Larger MMLU MMMLU MoE OpenAI OSS Qwen3 Uses VentureBeat

Provenance

Canonical issue: Ramsay Research Agent — 2026-03-15
AI generated: no
Story unit: 2026-03-15-qwen3-5-9b-beats-gpt-oss-120b-13x-larger-on-consumer-hardware
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Qwen3.5-9B Beats GPT-OSS-120B (13x Larger) on Consumer Hardware.

Story

Related stories

Moonshot cuts off new Kimi K3 signups after 48 hours

Kimi K2.7-Code: a 1T-param open coding model with benchmarks nobody can check

Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.

DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget

Meta ships Muse Glimmer, a 30B open agent model

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Qwen3.8-27B is architecturally byte-identical to Qwen3.6-27B. Every gain came from training.

Nemotron Cascade 2: NVIDIA's 30B-A3B MoE Outperforms Qwen3.5-35B

Source trail

Entities

Provenance

Qwen3.5-9B Beats GPT-OSS-120B (13x Larger) on Consumer Hardware.

Story

Related stories

Moonshot cuts off new Kimi K3 signups after 48 hours

Kimi K2.7-Code: a 1T-param open coding model with benchmarks nobody can check

Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.

DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget

Meta ships Muse Glimmer, a 30B open agent model

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Qwen3.8-27B is architecturally byte-identical to Qwen3.6-27B. Every gain came from training.

Nemotron Cascade 2: NVIDIA's 30B-A3B MoE Outperforms Qwen3.5-35B

Source trail

Entities

Provenance