DeepSWE results: GPT-6 Sol slightly below GPT-5.6 Sol, but cheaper

Melvin Vivas · X post · 2026-09-23 · Open on X

Topics: Industry Trends & Job Market, LLM Fundamentals · Level: intermediate

Summary

Shares DeepSWE benchmark results showing that GPT-6 Sol peaks slightly below its predecessor GPT-5.6 Sol on coding tasks, though it costs less. GPT-6 Luna is slightly better and cheaper than GPT-5.6 Luna, but it isn't a big jump. The lesson: a newer model version isn't automatically better, so compare benchmarks and cost for your own use case.

Key points

• On DeepSWE, GPT-6 Sol peaks slightly below GPT-5.6 Sol.
• GPT-6 Sol is cheaper than GPT-5.6 Sol.
• GPT-6 Luna improves slightly over GPT-5.6 Luna and is also cheaper.
• The generational gain is small, so a newer version doesn't guarantee better coding performance.
• When choosing a coding model, weigh benchmark scores against cost per task.

Resources mentioned

• DeepSWE · tool · deepswe.datacurve.ai · free
  A software-engineering benchmark used to compare how well coding models perform.
  Also in: DeepSeek V4 Flash 0731 Agentic Benchmarks and Use with Hermes Agent (Melvin Vivas on X · notes)
• Luna · tool · openai.com · paid
  The other model/agent the classifier routes tasks to; the post doesn't describe it further.
  Also in: GPT-6.1 Sol May Beat Luna for Subagents (Melvin Vivas on X · notes), Picking models for orchestrator and subagent roles in Codex (Melvin Vivas on X · notes), GPT-6 Sol Ultra Subagents Use Up Limits Fast (Melvin Vivas on X · notes), GPT-6 Sol vs Opus 5.5 in a Livestream Comparison (Melvin Vivas on X · notes) and 14 more
• GPT-6 Luna · tool · community.openai.com · paid
  An OpenAI model aimed at fast, high-volume classification and routing.
  Also in: DigitalOcean Serverless Inference Now Serves OpenAI GPT-6 Models (Melvin Vivas on X · notes), Codex Orchestrator v0.2.1: GPT-6 Sol orchestrator with Luna subagents (Melvin Vivas on X · notes), OpenAI Releases GPT-6 Sol and GPT-6 Luna (Melvin Vivas on X · notes)