LLM Reddit May 5, 2026 1 min read
DeepSeek V4 Pro tied with GPT-5.2 on FoodTruck Bench, a 30-day agentic benchmark using 34 tools, arriving roughly 10 weeks after GPT-5.2 was tested at approximately 17x lower cost.
DeepSeek V4 Ecosystem: From Open Weights to DeepClaude
Current state
Tracking the ripple effects of DeepSeek V4's open-weight release: near-frontier performance at a fraction of the cost, and how the community is already repurposing it in tools like DeepClaude.
What changed recently
- DeepSeek V4 Pro Matches GPT-5.2 on Agentic Benchmark — 17x Cheaper, 10 Weeks Later
- DeepClaude: Run Claude Code's Agent Loop with DeepSeek V4 Pro at 17x Less Cost
- DeepSeek V4: Near-Frontier LLM Performance at a Fraction of the Cost
Key tensions
Optimistic case: DeepSeek V4 Ecosystem: From Open Weights to DeepClaude unlocks real, compounding leverage.
Skeptical case: reliability, cost, and control around DeepSeek V4 Ecosystem: From Open Weights to DeepClaude remain unresolved.
Signals to watch
- Momentum and new coverage around “deepseek”
- Momentum and new coverage around “llm”
- Momentum and new coverage around “benchmark”
Timeline
Latest
Recent development
LLM Hacker News May 4, 2026 1 min read
DeepClaude keeps Claude Code's complete agent loop — file editing, bash, subagent spawning — while routing API calls to DeepSeek V4 Pro or other backends, cutting output token costs from $15/M to $0.87/M.
Recent development
LLM Hacker News May 2, 2026 1 min read
DeepSeek released DeepSeek-V4-Pro (1.6T total parameters, 49B active) and V4-Flash (284B total, 13B active), both Mixture-of-Experts models with MIT license and 1M token context. V4-Pro is the largest open-weights model released so far, and its pricing at $1.74/M input undercuts GPT-5.4 and Claude Sonnet 4.6 by more than half.