AI model

Amazon

2026-10-01 16:54:21

Amazon-backed Strand Labs releases Strands Decider 2B decision model

Strand Labs, an AI lab under Amazon Web Services, has released Strands Decider 2B, according to a Techub News brief citing TechCrunch. The new model is described as a decision model similar to Jevalike. The release comes as the number of decision models available online continues to rise. Techub said the model is intended to address that surge. No additional technical specifications, deployment details, or launch timeline beyond the release were disclosed in the brief.

50

NaiveAI

2026-09-28 03:40:43

NaiveAI open-sources first model as AI helps drive 151 optimization runs in six days

NaiveAI, founded by Tsinghua University associate professor Dai Jifeng, has released Naive-N0.5-Flash and opened its model weights under the MIT license. The model is based on Xiaomi’s MiMo-V2.5 Base and keeps the original mixture-of-experts, or MoE, architecture while changing the attention design. According to the team, global attention was replaced with sliding-window attention and DeepSeek sparse attention, followed by continued training on 3.25 trillion tokens. The updated model natively supports a 1 million-token context window, which the company said cuts compute costs for very long text processing. NaiveAI also said AI systems were directly involved in development, handling coding, experiments, result analysis and later optimization, while human researchers set direction and made key decisions. Using its NaiveRT inference system as an example, the team said it completed 151 optimization rounds over six days, with 63 adopted. The company also disclosed a peak inference speed of 2,122 tok/s under a specific setup, while saying standard mode runs at about 50 tok/s per user.

210

MiniMax

2026-09-27 13:01:44

MiniMax rolls out M3.1-Flash-Preview and adds limited-time perks on MiniMax Code

MiniMax has launched its new text model, M3.1-Flash-Preview, and made it available on MiniMax Code. The company said the model is built for everyday development work, with an emphasis on speed and stability. It is positioned for tasks such as fixing bugs quickly and handling full feature development. MiniMax Code is also offering two limited-time promotions. From Sept. 28 to Oct. 7 (UTC+8), users who check in daily can receive double free quota. The offer is open to both new and existing users, and the quota can be used on models including M3.1-Flash-Preview, H3, and H3 Max. MiniMax also said that after M3.1-Flash-Preview went live, all users’ Token Plan quota was reset immediately. According to the company, more quota resets will follow later.

220

Alibaba

2026-09-22 02:19:48

Market chatter says Alibaba plans AI model with 5 trillion to 10 trillion parameters

A brief item from ChainCatcher said market chatter indicates Alibaba is planning to train an artificial intelligence model with between 5 trillion and 10 trillion parameters. The note did not include product details, a development timeline, technical architecture, or any official confirmation from the company. No additional context on deployment, use cases, or release plans was provided in the source item. The report was presented as a market message rather than a formal corporate announcement.

320

xAI

2026-09-21 15:58:00

xAI launches Grok 4.7 with a focus on coding and knowledge work

xAI has released Grok 4.7, saying the model is built on a larger base model and trained for a longer period with reinforcement learning aimed at complex, long-duration tasks. The company said the update strengthens coding, long-context handling, and self-verification capabilities. Grok 4.7 is now available on Cursor, Grok Build, and via API. Pricing is set at $2 per million input tokens and $6 per million output tokens. xAI also said the new model shows clear gains over Grok 4.6 across multiple coding, office, and safety benchmarks. The announcement was cited by PANews on Sept. 21.

340

Qwen

2026-09-19 07:57:22

Qwen launches Qwen3.8-LiveTranslate with 2.3-second latency

Qwen has released Qwen3.8-LiveTranslate, a real-time simultaneous interpretation model that cuts LAAL, a metric used to measure average translation lag, to 2.3 seconds from 2.8 seconds in the previous generation. The update also adds real-time speaker separation, synchronized streaming output for source and translated text, and long-context disambiguation. According to the source report, the model can identify different speakers, align translated output with specific speakers, and keep translated speech more consistent with each speaker’s voice characteristics. It also uses prior context to resolve names, terminology, and pronoun references, reducing ambiguity that can arise when only the current sentence is considered. API pricing is largely unchanged from the prior version. Alibaba Cloud’s pricing page shows that in the Beijing region, audio input, image input, text output, and audio output for Qwen3.8-LiveTranslate are priced the same as Qwen3.5-LiveTranslate. In Singapore, prices for those four items are slightly lower.

280

Google

2026-09-12 09:31:53

Techub page gives no article details on Google’s TimesFM-3 model

A Techub page headline says Google has released TimesFM-3, a time-series forecasting model that can integrate sales and weather data. But the body text provided in the input does not contain any reporting details about the model itself. Instead, the page consists of infringement and compliance notices, contact information, a Techub News app download link, a general disclaimer, and a list of related article recommendations. Based on the supplied content, there are no verifiable details on the model’s release timing beyond the page timestamp, no technical specifications, no product description, and no additional context about how TimesFM-3 works. The source URL is a Techub link, and the published time in the input is 2026-09-12T09:31:53.503Z.

200

Techub

2026-09-11 12:42:22

Techub page does not include the reported details on Ling 3.0 Flash VL launch

The input identifies a Techub item titled "OpenRouter 上线蚂蚁集团旗下 inclusionAI 新模型 Ling 3.0 Flash VL" and lists it under Hot Articles, with a published timestamp of 2026-09-11T12:42:22.833Z. However, the body content provided in the input does not contain the underlying news report or any factual details about the claimed launch. What is actually visible in the supplied text is a standard page notice covering infringement and compliance reporting, contact information, a business cooperation email, a Techub News App download link, a website disclaimer, and several on-site recommended articles. The contact emails shown are creator@techub.news and contacting@techub.news, and the download URL listed is https://www.techub.news/download. Because the input does not include verifiable reporting on OpenRouter, Ant Group's inclusionAI, or the model Ling 3.0 Flash VL, no additional claims can be made here about the launch, model features, timing, product positioning, or technical specifications.

190