Trinity Large Thinking • Hacker News Board

Re: Trinity Large Thinking

#11

Post by kristianp » Thu, Apr 02, 2026, 4:50 AM UTC

edg5000 wrote: post #9

They are repeating a million times on their huggingface page that the thinking output should be included in the conversation history for multiturn use. That makes me wonder, is this generally needed for LLMs? Because that implies that they only really function well on typicial multiturn flows; I'm experimenting with a completely different approach: there is still the main message stream in the context, but the agent…

I've heard someone mention feeding back thinking when talking about gpt-oss-120, at the time that was the only evidence I could see that this is a thing.

Re: Trinity Large Thinking

#12

Post by anon373839 » Thu, Apr 02, 2026, 5:30 AM UTC

nl wrote: post #6

Trinity Large Preview managed 17/25 on my agentic SQL benchmark: https://sql-benchmark.nicklothian.com/?#all-data which is a fairly mediocre score for a large model (Qwen 27B managed 23/25) This non-Preview release scored 16/25. Probably the same model as the preview, or at least not particularly improved if you want agentic performance. Good to see more options for large open models though! It's hard to point defini…

Bit of a tangent, but I'm pleased to see that Qwen 3.5 35B is tied with GPT-5.4 and just 2 points behind 4.6 Opus. That little model is so impressively capable and fast! I'm frequently still surprised that I have that level of capability and speed running locally on my laptop.

Re: Trinity Large Thinking

#13

Post by nl » Thu, Apr 02, 2026, 5:38 AM UTC

anon373839 wrote: post #12

nl wrote: post #6

Trinity Large Preview managed 17/25 on my agentic SQL benchmark: https://sql-benchmark.nicklothian.com/?#all-data which is a fairly mediocre score for a large model (Qwen 27B managed 23/25) This non-Preview release scored 16/25. Probably the same model as the preview, or at least not particularly improved if you want agentic performance. Good to see more options for large open models though! It's hard to point defini…

Bit of a tangent, but I'm pleased to see that Qwen 3.5 35B is tied with GPT-5.4 and just 2 points behind 4.6 Opus. That little model is so impressively capable and fast! I'm frequently still surprised that I have that level of capability and speed running locally on my laptop.

Nemotron-Cascade-2-30B-A3B is worth checking out too at the size - I found it even better than Qwen 3.5 35B! I ran it slowly but successfully on a 8GB 1070GTX with CPU offload.

Re: Trinity Large Thinking

#14

Post by anon373839 » Thu, Apr 02, 2026, 5:41 AM UTC

nl wrote: post #13

anon373839 wrote: post #12

Earlier quoted context omitted.

Bit of a tangent, but I'm pleased to see that Qwen 3.5 35B is tied with GPT-5.4 and just 2 points behind 4.6 Opus. That little model is so impressively capable and fast! I'm frequently still surprised that I have that level of capability and speed running locally on my laptop.

Nemotron-Cascade-2-30B-A3B is worth checking out too at the size - I found it even better than Qwen 3.5 35B! I ran it slowly but successfully on a 8GB 1070GTX with CPU offload.