DeepSeek New Model Tested: Great Long-Context, But Tool Calling Quirks
TheZachMueller · x · 2026-08-01
After overnight testing of the suspected DeepSeek v4-flash (0731) model, a developer shared positive initial impressions. However, they noted that the prompt format and tool call/conversation history quirks could cause integration issues, potentially leading to mixed early reviews—though some negative feedback might stem from user or harness errors.
In a subsequent coding agent gauntlet, the tester found the model's task persistence so robust that they had to raise its token limit from 200k to over 300k before it would give up on a task.
More from coding & agent
- Founder wipes 30k lines of tech debt in one shot with Opus — lucasmeijer · 2026-10-16
- Japanese engineer runs an AI agent org inside tmux: per-project teams, one human talking to the orchestrator — 4310sy · 2026-10-16
- gloria-mcp ships an MCP server serving Gloria AI's curated crypto news — modelcontextprotocol · 2026-10-16
- Pine Script MCP server lets AI agents read v6 docs and validate syntax — modelcontextprotocol · 2026-10-16
- Open-source ebook2audiobook turns any ebook into a free audiobook, 20.4k GitHub stars — thisguyknowsai · 2026-10-16
- Composing music with math patterns via Gemini agents in Antigravity CLI — fofrAI · 2026-10-16