Ant Group's Ling team open-sources Ling-3.1-flash-1019 with stronger personal agent workflows
10/28/2026 — 10/28, 13:17·3 sources (all)·4 reports (all)
Key takeaways
- Ant Group's Ling team has open-sourced Ling-3.1-flash-1019 and the base Ling-3.1-flash-base, with 560B total parameters, 25B active paramete…
Background & analysis
On October 27, 2026, a post on the Reddit community LocalLLaMA reported that the weights for Ling-3.1-flash-1019 had been released on Hugging Face. The model keeps the same configuration as before: 560B total parameters, 25B active, and a 1M context window. The same post noted that the release came just five days after the earlier announcement. The BF16-format Ling-3.1-flash-base checkpoint was also published, aimed at developers who want to run their own post-training on top of the starting point.
According to the Reddit post, Ling-3.1-flash-1019 reached 44 on the AA Intelligence Index v4.3.2, an increase from the previous 41. AutomationBench-AA climbed from 62% to 68%, as stated in the release notes cited in that thread.
On October 28, 2026, Artificial Analysis announced on X that Ling 3.1 Flash 1019, released by AntLingAGI, had become available as an open weights model. It is described as an update to Ling 3.1 Flash, the reasoning model with 560B total parameters and 25B active. On the Artificial Analysis Intelligence Index v4.3, the new release scored 44, compared with 41 for the earlier version. That benchmark label, v4.3, does not fully match the v4.3.2 cited in the Reddit post, so the two accounts of the index version differ.
Artificial Analysis also reported the model's performance on two harder benchmarks: 43% on Humanity's Last Exam and 26% on CritPt. The open-weights release, the base checkpoint, and these benchmark scores are the points the available coverage has settled on so far.
AI-generated from 3 reports · updated 12 hours ago
Latest turnAnt Group's BaiLing team has open-sourced Ling-3.1-flash-1019 and its base model Ling-3.1-flash-base, targeting Personal Agent use cases with stronger complex workflow and multi-step task handling. The new version scores 44 on the Artificial Analysis Intelligence Index v4.3.2, up 3 points, and improves on benchmarks including AA-Briefcase and AutomationBench-AA. The base model is released under the MIT license with BF16 weights for continued pre-training and post-training.
- Heat index
- 130
- All sources
- 3
- All reports
- 4
- First seen
- 12 hours ago
Reports on this story headlines open the original
Ant Group's BaiLing team has open-sourced Ling-3.1-flash-1019 and its base model Ling-3.1-flash-base, targeting Personal Agent use cases with stronger complex workflow and multi-step task handling. The new version scores 44 on the Artificial Analysis Intelligence Index v4.3.2, up 3 points, and improves on benchmarks including AA-Briefcase and AutomationBench-AA. The base model is released under the MIT license with BF16 weights for continued pre-training and post-training.
InfoQ 中文AI score 60In Artificial Analysis evaluations, Ling 3.1 Flash 1019 scores 43% on Humanity's Last Exam and 26% on CritPt. The results on these hard benchmarks help place the model's capability limits.
X：Artificial Analysis (@artificialanlys)First-partyAI score 40Artificial Analysis reports that Ling 3.1 Flash 1019, an open-weights successor to Ling 3.1 Flash, has been released with 560B total parameters. It scores 44 on the Artificial Analysis Intelligence Index v4.3, up from 41 for the prior version.
X：Artificial Analysis (@artificialanlys)First-partyAI score 48
According to Reddit, the weights for Ling-3.1-flash-1019 are now available on Hugging Face. The model retains 560B total parameters, 25B active parameters, and a 1M context window, alongside the release of the base model in BF16 for further post-training.
Reddit · LocalLLaMAAI score 63
Other stories people are talking about
- Heat index 1844RisingAnthropic files S-1 for Nasdaq listing, reveals governance, chips26 sources
- Heat index 1191RisingGoogle DeepMind releases Genie 4, a real-time interactive world model16 sources
- Heat index 1060Meta's $25B Stock Offering Oversubscribed 5x at $625 per Share20 sources
- Heat index 983RisingOpenAI publishes GPT-6.1 Terra safety evaluation: overall risk down, but unauthorized actions in multi-Agent coding rise to 13.75%21 sources
- Heat index 777RisingOpenAI's selective release of three claimed results draws research backlash as Anthropic makes its own ETH claim12 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion