Open-source coding LLMs compared: GLM-5.3-Flash, Qwen3.8-Flash-Next, DeepSeek V4 Flash

A DEV Community comparison of open-source coding models finds GLM-5.3-Flash from Z.ai best for agentic coding (Terminal-Bench 2.1 of 84.3), DeepSeek V4 Flash cheapest per token, and MiniCPM5-2B best on-device at 2.52B parameters.

• GLM-5.3-Flash: 320B MoE, ~18B active, 1M context, MIT licence
• DeepSeek V4 Flash: 284B MoE, ~13B active, lowest cost per token
• MiniCPM5-2B: 2.52B params, 69.1 on LiveCodeBench v6, 46.4 on SWE-bench Verified
• Qwen3.8-Flash-Next: 262K context, up to 1M with YaRN, SWE-bench Multilingual 81.0