Tetsuyamin
@tetsuyamin.com
Grok
ideal for info retrieval. deep search is also great. sometimes helpful that the censorship is super lenient
Gemini
strong in all kinds of tasks, but it tends to glitch when generating long texts or code, and the ui is not user-friendly, so if you wanna use gemini, ai studio is sufficient
ideal for info retrieval. deep search is also great. sometimes helpful that the censorship is super lenient
Gemini
strong in all kinds of tasks, but it tends to glitch when generating long texts or code, and the ui is not user-friendly, so if you wanna use gemini, ai studio is sufficient
April 19, 2025 at 4:35 AM
Grok
ideal for info retrieval. deep search is also great. sometimes helpful that the censorship is super lenient
Gemini
strong in all kinds of tasks, but it tends to glitch when generating long texts or code, and the ui is not user-friendly, so if you wanna use gemini, ai studio is sufficient
ideal for info retrieval. deep search is also great. sometimes helpful that the censorship is super lenient
Gemini
strong in all kinds of tasks, but it tends to glitch when generating long texts or code, and the ui is not user-friendly, so if you wanna use gemini, ai studio is sufficient
Claude 3.7 Sonnet: greatest base model for coding
Quasar Alpha: OpenAI's promising experimental model for long context
Gemini 2.5 Pro Preview: latest SOTA model
Claude 3.7 Sonnet (thinking): greatest reasoning model for coding
GPT-4o-mini: literally worst on Aider Polyglot, not even tested much
Quasar Alpha: OpenAI's promising experimental model for long context
Gemini 2.5 Pro Preview: latest SOTA model
Claude 3.7 Sonnet (thinking): greatest reasoning model for coding
GPT-4o-mini: literally worst on Aider Polyglot, not even tested much
April 12, 2025 at 2:02 PM
Claude 3.7 Sonnet: greatest base model for coding
Quasar Alpha: OpenAI's promising experimental model for long context
Gemini 2.5 Pro Preview: latest SOTA model
Claude 3.7 Sonnet (thinking): greatest reasoning model for coding
GPT-4o-mini: literally worst on Aider Polyglot, not even tested much
Quasar Alpha: OpenAI's promising experimental model for long context
Gemini 2.5 Pro Preview: latest SOTA model
Claude 3.7 Sonnet (thinking): greatest reasoning model for coding
GPT-4o-mini: literally worst on Aider Polyglot, not even tested much
aside from 4o-mini the other 4 make total sense
April 12, 2025 at 1:58 PM
aside from 4o-mini the other 4 make total sense
Just to clarify upfront, I'm subscribed to ChatGPT Pro (the $200 tier), Claude Pro, SuperGrok, Gemini Advanced, and Perplexity Pro, and I also use OpenRouter to verify API specs.
March 21, 2025 at 3:58 PM
Just to clarify upfront, I'm subscribed to ChatGPT Pro (the $200 tier), Claude Pro, SuperGrok, Gemini Advanced, and Perplexity Pro, and I also use OpenRouter to verify API specs.
When it comes to coding, Claude 3.7 Sonnet Extended Thinking + Concise is overwhelmingly superior. For competitive programming tasks, o3-mini-high is powerful, though Claude is also quite strong. But for JavaScript, Claude is my absolute go-to.
March 21, 2025 at 3:58 PM
When it comes to coding, Claude 3.7 Sonnet Extended Thinking + Concise is overwhelmingly superior. For competitive programming tasks, o3-mini-high is powerful, though Claude is also quite strong. But for JavaScript, Claude is my absolute go-to.
For deep research purposes, I rely on Grok's DeeperSearch, with Gemini as a close second. Honestly, ChatGPT's Deep Research has been somewhat disappointing.
March 21, 2025 at 3:58 PM
For deep research purposes, I rely on Grok's DeeperSearch, with Gemini as a close second. Honestly, ChatGPT's Deep Research has been somewhat disappointing.
Qwen’s web service (Qwen Chat) also offers pretty decent OCR capabilities—but I wonder which model is running internally, considering Qwen Max isn't supposed to be multimodal.
March 21, 2025 at 3:58 PM
Qwen’s web service (Qwen Chat) also offers pretty decent OCR capabilities—but I wonder which model is running internally, considering Qwen Max isn't supposed to be multimodal.