English · 中文

GPT-5.5 with the Responses API: Reasoning Effort, Tools, and Long Tasks

May 8, 2026

Use a fixed GPT-5.5 snapshot in a controlled Responses API tool loop with reasoning policy, validation, idempotent execution, and recoverable checkpoints.

A First Look at GPT-5.5: Evaluating Capability for Real Work

Apr 27, 2026

Evaluate GPT-5.5 on frozen tasks, hidden acceptance, tool traces, blinded review, and total task cost after its April 2026 API availability.

Cross-Model Structured Output: Isolating Providers Behind One Domain Schema

Apr 18, 2026

Let OpenAI, Anthropic, and Gemini adapters own their structured-output parameters while mapping into one versioned domain schema and error model.

A Qwen3 Hybrid-Thinking Tutorial: Allocate Reasoning by Task

Apr 5, 2026

Turn Qwen3 hybrid thinking into a task router with bounded time and tokens, paired evaluation, deterministic verification, and explicit fallback.

A First Look at Mistral Small 4: Where Small Models Fit in Tool Use

Mar 25, 2026

Evaluate Mistral Small 4 for constrained tool routing while accounting for open weights, configurable reasoning, strict validation, and its real infrastructure floor.

  • « Prev
  • Page 5 of 14
  • Next »
Portrait of Lukes Lu

Lukes Lu

Software Developer

  • lukes.lu@yahoo.com
  • github.com/ilukes
  • @lupingui
  • Home
  • About
  • Keywords →
    • SwiftUI
    • OpenAI
    • Responses API
    • cancellation
    • iOS 26 beta
    • reasoning effort
    • AI Agent
    • Codex

©2026 All rights reserved. Made with Jekyll and ♥