alex spînu | lists

Favorite LLMs

11th of June 2026Tier list12 items

The ones in the tiers S and A are the ones I actively use.

S

  • GPT-5.6 Terra(OpenAI) — My daily driver and go-to model for coding and reasoning tasks. I rock the $100 subscription and the intelligence it provides for the dollar are what fits me(and my budget) the best.coding / main / openai
  • GPT-5.6 Sol(OpenAI) — Intelligence and reasoning powerhouse, whenever Terra cannot handle it.coding / main / openai
  • Claude Fable 5(Anthropic) — Same as GPT 5.6 except I can't get as much done in the $100 tier.coding / main / anthropic

A

  • Claude Opus 4.8(Anthropic) — Like GPT-5.5, but with a slower inference time and the limits are not sufficient for my budget. Minus point is that I feel like I'm extra careful at my prompts so the system does not mistakenly flag me. The frontend bits is what stands out when using Claude.coding / frontend

B

  • Kimi 2.6(fast coding utility) — My favorite open model for coding and reasoning tasks. Fast and reliable, with a good balance of performance and cost. Great at frontend.coding / fast / frontend / reliable
  • GLM 5.2(deep reasoning) — Great at in-depth reasoningreasoning / glm / in-depth
  • Sonnet 5(Anthropic) — Feels weird to put it here but the cost per intelligence doesn't cut it for me when I can get it done with GPT 5.5 at more than half the cost or even Fable 5.general / sonnet / anthropic

C

  • Gemini 3.5 Flash(web chat companion) — My main model I use as chat model within the Gemini web appchat / gemini / web-app
  • Qwen 3.7 Max(experimental favorite) — Qwen models's personality is one that I always resonated with, I use it a lot in experiments along with it's younger siblings Qwen 3.5experimental / qwen-family / personality
  • DeepSeek V4 Flash(cost-effective workhorse) — This one is in the same tier as the Pro variant because it's a very capable model for the incredible price, tool usage, and speed. I use it extensively for things like summarization and custom workflows like asking it questions about my new emailsfast / tool-use / summarization / workflows / affordable

D

  • Deepseek V4 Pro(planning specialist) — Good at planningplanning / deepseek / architecture
  • Mimo v2.5 Pro(no-frills utility) — Like Kimi but it lacks a personality, I couldn't pinpoint what exactly it is but it just felt blunt.utility / minimal