DeepSeek V4.1 Flash Snapshot behebt echten Bug in Claude Code und OpenCode
Der 17.-Oktober-Snapshot von DeepSeeks V4.1 Flash laesst sich nun unter der festen ID deepseek-v4.1-flash-1017 aufrufen. Das Pinning ist fuer Agenten wichtig: Das getestete Verhalten bleibt erhalten. Der Beitrag zeigt ausserdem, was die einzelnen Laeufe gekostet haben.
Posted on Oct 27 • Originally published at xhuoapi.ai
DeepSeek's October 17 snapshot of V4.1 Flash is the model behind the deepseek-flash name on DeepSeek's API, and it is now callable under its pinned ID deepseek-v4.1-flash-1017 on XHuoAPI. Pinning the snapshot matters for agents: the behaviour you tested is the behaviour you keep.
Disclosure: I run XHuoAPI. DeepSeek's own prices are listed next to ours, with the source, so you can compare.
A tiny Node project with one bug: orders of $50 or more should have no shipping fee, but total > 50 charges a $50 order. Each tool got the same prompt, "The test in shipping.test.js fails. Find the bug in shipping.js and fix it", and a run counted only if node --test passed afterwards.
Every run found the > vs >= mistake and made a one-character edit.
Claude Code sent about 58k–71k input tokens per run (most of it Claude Code's own system prompt and tool definitions) and got 550–750 output tokens back. At XHuoAPI's Standard price that is about $0.02 per run at peak hours and about half that off-peak, before any cache discount.
DeepSeek bills peak hours (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday) at twice the off-peak rate. XHuoAPI uses the same hours for DeepSeek models.
Cheaper tiers trade stability for price. The same key also calls Qwen, GLM, Kimi, GPT and Claude.
Claude Code (leave /v1 off the base URL):
export ANTHROPIC_BASE_URL = https://api.xhuoapi.ai export ANTHROPIC_AUTH_TOKEN = YOUR_KEY export ANTHROPIC_MODEL = deepseek-v4.1-flash-1017 export ANTHROPIC_DEFAULT_HAIKU_MODEL = deepseek-v4.1-flash-1017 Enter fullscreen mode Exit fullscreen mode OpenCode 2, opencode.jsonc:
{ "$schema": "https://opencode.ai/config.json", "providers": { "xhuoapi": { "name": "XHuoAPI", "env": [ "XHUOAPI_API_KEY" ], "package": "@opencode/ai/providers/openai-compatible", "settings": { "baseURL": "https://api.xhuoapi.ai/v1" }, "models": { "deepseek-v4.1-flash-1017": { "name": "DeepSeek V4.1 Flash 1017" } } } } } Enter fullscreen mode Exit fullscreen mode Run it with opencode run -m xhuoapi/deepseek-v4.1-flash-1017 "...".
Thinking is on by default; thinking: {"type": "disabled"} turns it off on our endpoint. Batch jobs are cheaper off-peak on both DeepSeek and XHuoAPI.
Full comparison (DeepSeek, DeepInfra, OpenRouter, XHuoAPI): https://xhuoapi.ai/en/guides/deepseek-api-outside-china.html?utm_source=devto&utm_medium=article&utm_campaign=deepseek-1017 Per-tool test results: https://xhuoapi.ai/en/guides/chinese-models-in-coding-agents.html?utm_source=devto&utm_medium=article&utm_campaign=deepseek-1017