Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. Scored as launched (GLM-4.6V, the 106B flagship, 8 Dec 2025; GLM-4.6V-Flash is the 9B tier, free via Z.ai API, FlashX $0.04/$0.40). Reasoning null: no Artificial Analysis Intelligence Index entry found; Z.ai's headline is SoTA or near-SoTA among open models of its size on 20+ vision benchmarks. Coding null. Context 128K tokens (trained, Z.ai docs) -> 1. Price $0.30 in / $0.90 out per 1M (Z.ai pricing page; OpenRouter) = $0.45 blended -> 5. Speed null. Modalities: video, image, text and file input, text output (Z.ai docs) -> 4. Reviews: VentureBeat and MarkTechPost were Cloudflare/Vercel-blocked, so Wayback snapshots (opened this run) are cited; SiliconFlow's launch blog is a reseller post. No independent hands-on test found; all three repeat Z.ai's benchmark claims. (A review whose quote could not be re-verified was dropped.)[2]
It matches or outperforms comparable-scale models like Qwen3-VL-235B , Kimi-VL-A3B-Thinking-2506 , and Step3-321B in key capabilities: multimodal understanding, multimodal agentic tasks, and long-context processing.
The 106B model’s 128K-token window allows it to outperform larger models like Step-3 (321B) and Qwen3-VL-235B on long-context document tasks, video summarization, and structured multimodal reasoning.
The model consumes those visual outputs and fuses them with text in the same reasoning chain. This closes the loop from perception to understanding to execution and is explicitly positioned as the bridge between visual perception and executable action for multimodal agents.
| Z.ai debuts open source GLM-4.6V, a native tool-calling vision model for multimodal reasoning · 8 de dez. de 2025 | 0.50 |
|---|---|
| Zhipu AI Releases GLM-4.6V: A 128K Context Vision Language Model with Native Tool Calling - MarkTechPost · 9 de dez. de 2025 | 0.85 |
| GLM-4.6V Now on SiliconFlow: Native Multimodal Tool Use Meets SoTA Visual Intelligence · 11 de dez. de 2025 | 0.60 |
Fotos de leitores, separadas das mídias com fontes do pin. Avalie o que você comeu ou viu e, em um pin de restaurante, marque os pratos direto na sua foto.
A primeira entrada é sempre a fonte do pin. A confiança geral é uma média ponderada de quanto cada referência sustenta os horários de início e fim usados acima; uma referência conta metade a cada 180 dias a mais em relação à mais recente.
A VentureBeat[4] informou em 8 de dezembro de 2025 (5:03 pm PT) que a Z.ai "has released its GLM-4.6V series" e a OpenRouter[5] lista z-ai/glm-4.6v em 8 de dezembro; os repositórios do Hugging Face foram enviados em 7 de dezembro UTC e os vídeos de estudos de caso do lançamento em docs.z.ai[2] estão rotulados como 1208.
Z.ai developer docs: GLM-4.6V, GLM-4.6V-FlashX and GLM-4.6V-Flash, 128K context, native multimodal tool use, image-text content creation, visual web search, frontend replication, about 150 pages, 200 slides or one hour of video in one pass, more than 20 benchmarks.[1]
GLM-V paper, revised 1 January 2026 (v6): now covers GLM-4.1V-Thinking, GLM-4.5V and GLM-4.6V, a reasoning-centric RL recipe (RLCS) and the GLM-4.6V series adding native tool use and a 128K-token window; open-sourced models.[1]
VentureBeat (Carl Franzen), 8 Dec 2025: Z.ai released the open-source GLM-4.6V series (106B and 9B Flash) with native function calling, a 128K-token context, MIT license, $0.30 and $0.90 per million tokens and free Flash; use cases are frontend automation, long documents and video summarization.
OpenRouter listing (read from its models API): Z.ai: GLM 4.6V, added 8 Dec 2025, 131,072-token context, image, text and video input, $0.30 per million input and $0.90 per million output tokens, up to 32,768 output tokens.
Falta algo ou há um erro? Diga com suas palavras: um link que sustente este pin, uma data de início ou fim diferente e por quê, ou um fato que falta ou está errado. A IA confere isso com as fontes do pin, procura outras melhores e adiciona qualquer página que dê razão a você. As fontes do próprio pin continuam contando mais. Uma imagem que mostra outra coisa, ou mostra mal, também é analisada e rebaixada ou substituída.
Título original: GLM-4.6V and GLM-4.6V-Flash Released
Um subconjunto de linhas do gráfico de resultados do GLM-4.6V da Z.ai (uma imagem no cartão do modelo e em docs.z.ai), modelos de raciocínio; "/" significa não informado[1].
| Benchmark | GLM-4.6V (106B) | GLM-4.6V-Flash (9B) | GLM-4.5V (106B) | Qwen3-VL-8B-Thinking | Qwen3-VL-235B-A22B-Thinking |
|---|---|---|---|---|---|
| MMBench V1.1 | 88.8 | 86.9 | 88.2 | 84.3* | 90.6 |
| MMStar | 75.9 | 74.7 | 75.3 | 75.3 | 78.7 |
| MMMU (Val) | 76.0 | 71.1 | 75.4 | 74.1 | 80.6 |
| MMMU-Pro | 66.0 | 60.6 | 65.2 | 60.4 | 69.3 |
| MathVista | 85.2 | 82.7 | 84.6 | 81.4 | 85.8 |
| AI2D | 88.8 | 89.2 | 88.1 | 84.9 | 89.2 |
| Design2Code | 88.6 | 69.8 | 82.2 | 56.6* | 93.4 |
| OSWorld | 37.2 | 21.1 | 35.8 | 33.9 | 38.1 |
| AndroidWorld | 57.0 | 42.7 | 57.0 | 50.0 | / |
| WebVoyager | 81.0 | 71.8 | 84.4 | 47.7* | 30.4* |
| MMLongBench-Doc | 54.9 | 53.0 | 44.7 | 48.0 | 56.2 |
| OCRBench | 86.5 | 84.7 | 86.5 | 81.9 | 87.5 |
| ChartQAPro | 65.5 | 62.6 | 64.0 | 58.4* | 63.6* |
| RefCOCO-avg (val) | 88.6 | 85.6 | 91.3 | 89.3* | 92.4 |
| Ref-L4-test | 88.9 | 87.7 | 89.5 | 88.6* | 90.4 |