--- title: 'Chinese Open-Weight Coding LLMs: 2026''s Three-Week Sweep' description: In 17 days, GLM-5.1, Kimi K2.6, and DeepSeek V4 shipped frontier-tier open-weight coding LLMs at a fraction of Western prices. Inside the April 2026 wave. url: https://nerdleveltech.com/chinese-open-weight-coding-llms-three-week-sweep language: en date: '2026-05-11' category: ai-ml tags: - open-weight LLM - open-source coding LLM - Chinese AI models - GLM-5.1 - Kimi K2.6 - DeepSeek V4 - Z.ai - Moonshot AI - DeepSeek - SWE-Bench Pro - SWE-bench Verified - LiveCodeBench - Codeforces - Claude Opus 4.7 - GPT-5.5 - MIT license - Mixture of Experts - agentic coding - long-horizon coding - open-weight comparison - frontier AI markdown_url: https://nerdleveltech.com/chinese-open-weight-coding-llms-three-week-sweep.md attribution: 'Source: NerdLevelTech — https://nerdleveltech.com/chinese-open-weight-coding-llms-three-week-sweep' terms: https://nerdleveltech.com/llms.txt --- # Chinese Open-Weight Coding LLMs: 2026's Three-Week Sweep ## TL;DR Between **April 7 and April 24, 2026**, three Chinese labs shipped open-weight coding models that closed almost the entire gap to the Western frontier. **GLM-5.1** (Z.ai, April 7) became the first open-weight model to top SWE-Bench Pro at **58.4%**, beating GPT-5.4's 57.7%.[^1][^2] Nine days later **Claude Opus 4.7** (Anthropic, April 16) reclaimed the lead at **64.3%** and pushed SWE-bench Verified to **87.6%**.[^3][^4] **Kimi K2.6** (Moonshot AI, April 20) answered with **58.6% on SWE-Bench Pro and 80.2% on SWE-bench Verified** — a 1-trillion-parameter MoE that orchestrates up to **300 sub-agents** in a single coding run.[^5][^6] Four days later, **DeepSeek V4 Preview** (April 24) shipped a 1.6T MoE with **1M-token context, 80.6% SWE-bench Verified, a 3,206 Codeforces rating** (above GPT-5.4's 3,168), and **MIT-licensed weights** at roughly **one-seventh the output cost** of Claude Opus 4.7.[^7][^8][^9] None of the three close the gap to Opus 4.7 on SWE-Bench Pro — but on price-per-task, agent orchestration, context length, and license clarity, the open-weight frontier is now within touching distance of the closed one. --- ## What You'll Learn - The exact 17-day timeline of releases — and why the order matters - How each model scores on SWE-Bench Pro, SWE-bench Verified, LiveCodeBench, and Codeforces - The price gap between the three Chinese open-weight models and GPT-5.5 / Claude Opus 4.7 - Why MIT vs Modified MIT vs non-commercial licensing changes who can actually use these weights - The three different bets each lab made — and which one fits your workload - Where the remaining gap to Claude Opus 4.7 still hurts, and where it doesn't --- ## A 17-Day Window That Changed the Open-Weight Frontier Before April 7, the open-weight coding leaderboard sat well behind the closed frontier. SWE-Bench Pro's top open-weight entrant trailed GPT-5.4 by several points. SWE-bench Verified was a wider gap. Then three releases landed in seventeen days. | Date | Release | Lab | Headline result | |---|---|---|---| | **April 7, 2026** | GLM-5.1 | Z.ai (formerly Zhipu) | First open-weight #1 on SWE-Bench Pro at **58.4%**[^1][^2] | | **April 16, 2026** | Claude Opus 4.7 | Anthropic | Reclaims SWE-Bench Pro lead at **64.3%**, ships **87.6%** SWE-bench Verified[^3][^4] | | **April 20, 2026** | Kimi K2.6 | Moonshot AI | **58.6%** SWE-Bench Pro, **80.2%** SWE-bench Verified, 300-agent swarm[^5][^6] | | **April 23, 2026** | GPT-5.5 | OpenAI | **58.6%** SWE-Bench Pro (tied with Kimi K2.6) and **88.7%** SWE-bench Verified — takes Verified lead from Opus 4.7[^10][^18] | | **April 24, 2026** | DeepSeek V4 Preview | DeepSeek | 1.6T MoE, **80.6%** SWE-bench Verified, **3,206** Codeforces, 1M context[^7][^8][^9] | The story isn't that any one Chinese model surpassed Anthropic's flagship. None of them did on SWE-Bench Pro. The story is that within three weeks, the *open-weight* frontier matched or beat *every other closed frontier vendor* — including GPT-5.5 — on at least one major coding benchmark, while undercutting them on price by 6× to 30×. --- ## SWE-Bench Pro: The Hardest Public Coding Benchmark SWE-Bench Pro evaluates an agent's ability to resolve real GitHub issues end-to-end. It is harder than SWE-bench Verified because it includes long-horizon and multi-file changes. Here is where the field stands today. | Model | SWE-Bench Pro | Released | License | |---|---|---|---| | Claude Opus 4.7 | **64.3%** | April 16, 2026 | Closed[^4] | | Kimi K2.6 | 58.6% | April 20, 2026 | Modified MIT[^5][^6] | | GPT-5.5 | 58.6% | April 23, 2026 | Closed[^10] | | GLM-5.1 | 58.4% | April 7, 2026 | MIT[^1][^2] | | GPT-5.4 | 57.7% | March 5, 2026 | Closed[^1] | | DeepSeek V4-Pro (max) | 55.4% | April 24, 2026 | MIT[^7][^8] | | Claude Opus 4.6 | 53.4% | February 5, 2026 | Closed[^6] | Two things stand out. First, the gap from the open-weight leader (Kimi K2.6 at 58.6%) to the closed leader (Opus 4.7 at 64.3%) is **5.7 percentage points** — wider than the gap between GPT-5.5 and Kimi K2.6 (zero). Second, GPT-5.5 and Kimi K2.6 are *tied* on this benchmark, but GPT-5.5 costs **$5 per million input tokens and $30 per million output**, while Kimi K2.6 lists at **$0.95 input / $4.00 output** on Moonshot's API.[^11][^12][^14] If you score on SWE-Bench Pro alone, the open-weight frontier has matched GPT-5.5 at a fraction of the price. The remaining gap is to Anthropic's Opus 4.7 — and that gap is real but narrow enough that for many workloads it does not justify a 6-30× output-token premium. --- ## SWE-bench Verified, LiveCodeBench, and Codeforces SWE-Bench Pro is the headline benchmark, but it's not the only one. Different models win different categories. | Model | SWE-bench Verified | LiveCodeBench | Codeforces Elo | |---|---|---|---| | GPT-5.5 | **88.7%** (OpenAI)[^18] | Not headline-reported | Not officially reported | | Claude Opus 4.7 | 87.6%[^4] | 89.6%[^4] | Not officially reported | | DeepSeek V4-Pro (max) | 80.6%[^7][^8] | **93.5%**[^7] | **3,206**[^7][^8] | | Kimi K2.6 | 80.2%[^5] | Not headline-reported | Not headline-reported | | Claude Opus 4.6 | 80.8%[^7] | 88.8%[^7] | Not officially reported | DeepSeek V4-Pro wins LiveCodeBench (93.5%) and Codeforces (3,206 — ahead of GPT-5.4's 3,168). Kimi K2.6 sits within a fraction of a point of Opus 4.6 on SWE-bench Verified. GPT-5.5 narrowly leads SWE-bench Verified at 88.7% (OpenAI-reported), with Claude Opus 4.7 close behind at 87.6%. Both lead the open-weight frontier (DeepSeek V4-Pro at 80.6%) by roughly 7 to 8 points on that benchmark. Note that vendor-reported SWE-bench Verified numbers use each lab's own scaffold and tool budget; on the independent vals.ai leaderboard, which fixes a neutral scaffold, GPT-5.5 lands at ~82.6%, closer to the open-weight pack.[^13] Apples-to-apples comparison across labs is harder than the headline numbers suggest. The pattern: open-weight models are competitive on competitive-programming and code-generation tasks (LiveCodeBench, Codeforces) and within striking distance on issue-resolution tasks (SWE-bench Verified). Long-horizon agentic resolution (SWE-Bench Pro) remains the area where Anthropic holds the clearest lead. --- ## The Price Gap That Actually Matters Per-token pricing tells the story of why these releases are reshaping production deployments. | Model | Input ($/M) | Output ($/M) | Notes | |---|---|---|---| | Claude Opus 4.7 | $5.00 | $25.00 | Closed[^11] | | GPT-5.5 | $5.00 | $30.00 | Closed; nearly doubled vs GPT-5.4 ($2.50/$15)[^12] | | Kimi K2.6 (Moonshot API) | $0.95 | $4.00 | Modified MIT, self-host available[^14] | | DeepSeek V4-Pro (list) | $1.74 | $3.48 | MIT[^9] | | DeepSeek V4-Pro (promo) | $0.435 | $0.87 | 75% promo until **May 31, 2026**[^9] | | GLM-5.1 (OpenRouter) | $1.05 | $3.50 | MIT, varies by provider[^15] | On output cost alone: - **Kimi K2.6 is ~6× cheaper than Claude Opus 4.7 and ~7.5× cheaper than GPT-5.5.** - **DeepSeek V4-Pro (list) is ~7× cheaper than Opus 4.7 and ~8.6× cheaper than GPT-5.5.** - **DeepSeek V4-Pro (promo) is ~29× cheaper than Opus 4.7 — until May 31.** And these are inference-API prices. The MIT-licensed weights mean you can self-host either DeepSeek V4 or GLM-5.1 with zero per-token cost. For a heavy coding workload that emits millions of output tokens per day, the math gets uncomfortable fast for closed-source incumbents. --- ## Three Different Bets The three Chinese labs did not all build the same kind of model. Each made a distinct architectural and product bet. ### GLM-5.1: The MIT-Licensed Long-Horizon Bet GLM-5.1 is a **754-billion-parameter MoE with 40B active per token**, MIT-licensed, with a **200K context window and 128K maximum output**.[^16] Z.ai's pitch is **long-horizon agentic coding**: the model is engineered to run autonomously for up to **eight hours of continuous task execution** without human checkpoints.[^2] It was trained on **100,000 Huawei Ascend 910B chips** — a deliberate signal that Chinese labs can train frontier models without NVIDIA hardware.[^2] Best for: agentic engineering workflows where the model needs to plan, execute, verify, and revise across multiple files and sub-tasks without supervision. ### Kimi K2.6: The 300-Agent Swarm Bet Kimi K2.6 is a **1-trillion-parameter MoE with 32B active per token**, Modified MIT licensed, with a **256K context window**.[^5][^6] Moonshot's headline architectural feature is **Agent Swarm**: a single coding run can orchestrate up to **300 sub-agents executing 4,000 coordinated steps**.[^6] Moonshot demonstrated 13 hours of continuous autonomous coding on an open-source financial matching engine during the launch.[^5] Best for: workloads that decompose naturally into parallel sub-tasks — refactoring across hundreds of files, test generation, multi-language porting. ### DeepSeek V4: The Architecture-Driven Cost Bet DeepSeek V4 ships as **two models**: V4-Pro (1.6T total / 49B active) and V4-Flash (284B total / 13B active), both under **MIT license** with a **1-million-token context window** and **384K max output**.[^8][^9] The architectural headline is **Hybrid Attention** — Compressed Sparse Attention and Heavily Compressed Attention layers that cut per-token inference FLOPs to **27% of V3.2's** and KV cache to **10%** at 1M-token context.[^8] V4 is the only model in the wave with a million-token window. Best for: large-codebase work where context length is the binding constraint — entire repositories, long debugging sessions, multi-document code review. --- ## License Differences That Matter in Production The "open" in "open-weight" varies meaningfully across the three. | Model | License | Commercial use | Self-host | |---|---|---|---| | GLM-5.1 | **MIT** | Yes, unrestricted[^16] | Yes | | DeepSeek V4 | **MIT** | Yes, unrestricted[^8] | Yes | | Kimi K2.6 | **Modified MIT** | Yes — attribution required for deployments >100M MAU or >$20M monthly revenue[^14] | Yes | | MiniMax M2.7 (earlier, March 18) | Restrictive Modified-MIT | "Written authorization required for commercial use"[^17] | Allowed for non-commercial | GLM-5.1 and DeepSeek V4 shipping under **standard MIT** is the legal end-state most enterprise buyers want. Kimi K2.6's Modified MIT functions as standard MIT for the vast majority of teams — its one added clause requires prominent "Kimi K2" attribution only for deployments exceeding 100 million monthly active users or $20 million in monthly revenue.[^14] MiniMax M2.7 — released earlier on March 18, 2026, ahead of the April wave — illustrates the pattern that has *not* taken hold: a "modified MIT" with non-commercial restrictions that requires prior written authorization for any commercial use.[^17] For enterprise procurement, the practical hierarchy is: GLM-5.1 ≈ DeepSeek V4 (no friction) → Kimi K2.6 (light review) → MiniMax M2.7 (separate license). --- ## What Each Model Actually Wins At Comparing the three Chinese open-weight models head-to-head — and against the closed frontier where relevant: - **Best open-weight SWE-Bench Pro:** Kimi K2.6 (58.6%), narrowly ahead of GLM-5.1 (58.4%); also ties closed-source GPT-5.5 (58.6%) - **Best LiveCodeBench (vendor-reported):** DeepSeek V4-Pro (93.5%) - **Best Codeforces:** DeepSeek V4-Pro (3,206 Elo) - **Best long-horizon autonomous coding:** GLM-5.1 (8-hour continuous task execution) - **Best parallel sub-task orchestration:** Kimi K2.6 (300-agent swarm) - **Best large-codebase context (among the three):** DeepSeek V4 (1M tokens) - **Best price per output token:** DeepSeek V4-Pro promo (until May 31) - **Cleanest license for enterprise procurement:** GLM-5.1 or DeepSeek V4 (standard MIT) The gap to Claude Opus 4.7 on SWE-Bench Pro (5.7 points) and SWE-bench Verified (~7 points) remains real. For mission-critical autonomous coding agents where every percentage point of issue-resolution accuracy matters, Opus 4.7 still has the lead — at 6× to 30× the output-token cost depending on which open-weight model and pricing tier you compare it to. --- ## The Gap That Remains It would be wrong to say the open-weight frontier has caught the closed frontier. Claude Opus 4.7's 64.3% on SWE-Bench Pro is the highest result publicly reported on that benchmark, and on SWE-bench Verified the closed frontier — GPT-5.5 at 88.7% and Opus 4.7 at 87.6% — still sits roughly 7-8 points above the open-weight leaders. On long-horizon agentic tasks involving multi-file resolution under ambiguity, Opus 4.7 currently wins. But on most coding tasks that are not at the absolute frontier of difficulty, the open-weight models are now within statistical noise of GPT-5.5 — and dramatically cheaper. The interesting question is no longer "can open-weight catch the closed frontier?" but "for which subset of coding workloads does the closed-frontier premium still pay for itself?" For high-stakes one-shot agentic resolution — yes. For bulk code generation, refactoring, test writing, multi-file editing, and CI-integration loops — increasingly, no. --- ## FAQ **Which Chinese open-weight model is best for coding in 2026?** It depends on the workload. Kimi K2.6 leads on SWE-Bench Pro at 58.6%, narrowly ahead of GLM-5.1 (58.4%) and well ahead of DeepSeek V4-Pro (55.4%). DeepSeek V4-Pro leads on LiveCodeBench (93.5%) and Codeforces (3,206 Elo). GLM-5.1 leads on long-horizon autonomous execution. No single model wins every benchmark, so the deciding factors are licensing, context length, price, and which kind of coding workload you optimize for. **Are these models truly open source?** They are **open weight**, not strictly open source. All three publish weights on Hugging Face. None publishes the full training dataset or training pipeline source. GLM-5.1 and DeepSeek V4 use standard MIT license; Kimi K2.6 uses a Modified MIT license. **How do they compare to Claude Opus 4.7?** On SWE-Bench Pro, Opus 4.7 leads the best open-weight model (Kimi K2.6) by 5.7 percentage points. On SWE-bench Verified, Opus 4.7 leads by 7 points. On price, Opus 4.7 costs roughly 6× more per output token than Kimi K2.6 and 7× more than DeepSeek V4-Pro list pricing. **Can I self-host these models?** Yes — all three publish weights on Hugging Face. Practical hosting requires substantial GPU resources (the active-parameter counts are smaller than the totals, but you still need to keep all expert weights in memory or accept routing-driven memory swaps). Most production users access them via API rather than self-hosting. **Why did three frontier-tier Chinese coding models ship within 17 days?** The timing is partly coincidence and partly competitive pressure — Claude Opus 4.7 launched on April 16 between Z.ai's release on April 7 and Moonshot/DeepSeek's later-April releases. Each lab had been working on its next-generation model for months; the convergence in release dates reflects similar training-cycle timelines, not coordination. --- ## Bottom Line In a 17-day window from April 7 to April 24, 2026, three Chinese open-weight coding models — GLM-5.1, Kimi K2.6, and DeepSeek V4 — closed almost the entire benchmark gap to GPT-5.5 while undercutting GPT-5.5 and Claude Opus 4.7 on price by 6× to 30×. Anthropic's Claude Opus 4.7 still leads on the hardest agentic coding benchmark (SWE-Bench Pro at 64.3%), but for the broader class of coding workloads, MIT-licensed open-weight models with 200K-to-1M-token context windows are now a credible production default. The 2026 frontier of coding LLMs is not a single line. It is a cluster — and the open-weight side of that cluster has, for the first time, become competitive on every dimension except the very hardest. --- ## Related reads - [Kimi K3 vs Fable 5, GPT-5.6, Grok: 2026 Benchmarks](https://nerdleveltech.com/kimi-k3-benchmarks-vs-fable-5-gpt-5-6-grok) [^1]: Winbuzzer, *Z.ai Releases GLM-5.1: 754B Model Tops SWE-Bench Pro*, April 9, 2026. https://winbuzzer.com/2026/04/09/z-ai-releases-glm-5-1-754b-model-tops-swe-bench-pro-xcxwbn/ [^2]: NerdLevelTech, *GLM-5.1: The Open-Source Model That Beat GPT-5.4*, April 19, 2026. /glm-5-1-open-source-beats-gpt-coding-benchmarks [^3]: TheNextWeb, *Claude Opus 4.7 leads on SWE-bench and agentic reasoning*, April 16, 2026. https://thenextweb.com/news/anthropic-claude-opus-4-7-coding-agentic-benchmarks-release [^4]: LLM-Stats, *Claude Opus 4.7 Benchmarks, Pricing & Context Window*. https://llm-stats.com/models/claude-opus-4-7 [^5]: Kimi.com Blog, *Kimi K2.6 Tech Blog: Advancing Open-Source Coding*. https://www.kimi.com/blog/kimi-k2-6 [^6]: NerdLevelTech, *Kimi K2.6: Open-Weight 300-Agent Swarm Tops GPT-5.4*, April 27, 2026. /kimi-k2-6-300-agent-swarm-open-weight-frontier-coding [^7]: DeepSeek API Docs, *DeepSeek V4 Preview Release*, April 24, 2026. https://api-docs.deepseek.com/news/news260424 [^8]: NerdLevelTech, *DeepSeek V4: Open-Weight Frontier at 1/7 the Cost*, May 2, 2026. /deepseek-v4-open-source-frontier-million-token-context [^9]: DeepSeek API Docs, *Models & Pricing*. https://api-docs.deepseek.com/quick_start/pricing/ [^10]: OpenAI, *Introducing GPT-5.5*, April 23, 2026. https://openai.com/index/introducing-gpt-5-5/ [^11]: Anthropic Pricing. https://www.anthropic.com/claude/opus [^12]: OpenAI API Pricing. https://openai.com/api/pricing/ [^13]: vals.ai SWE-bench leaderboard, May 2026. [^14]: Moonshot AI on Hugging Face, *Kimi-K2.6 Model Card*. https://huggingface.co/moonshotai/Kimi-K2.6 [^15]: OpenRouter, *GLM-5.1 API Pricing & Benchmarks*. https://openrouter.ai/z-ai/glm-5.1 [^16]: LLM-Stats, *GLM-5.1 Benchmarks, Pricing & Context Window*. https://llm-stats.com/models/glm-5.1 [^17]: MarkTechPost, *MiniMax Just Open Sourced MiniMax M2.7*, April 12, 2026. https://www.marktechpost.com/2026/04/12/minimax-just-open-sourced-minimax-m2-7-a-self-evolving-agent-model-that-scores-56-22-on-swe-pro-and-57-0-on-terminal-bench-2/ [^18]: TokenMix, *GPT-5.5 Review: 88.7% SWE-Bench, 92.4% MMLU, 2x Price Tag (2026)*. https://tokenmix.ai/blog/gpt-5-5-spud-review-88-swe-bench-2026 ; OpenAI announcement, April 23, 2026.