# Kimi K3 Is the Open-Model Champion in Waiting **Plutonous** | July 17, 2026 | 9 min read > Kimi K3 pairs a 57.1 Artificial Analysis score with 2.8T parameters, 1M-token context, $0.94 task cost, and open weights promised for July 27. Tags: Kimi K3, Moonshot AI, Open-Weight AI, Artificial Analysis, AI Benchmarks, Coding Agents, Model Economics, Chinese AI --- **TL;DR:** Kimi K3 scores **57.1** and ranks **#4 of 189** in its current Artificial Analysis comparison class, just **1.8 points** behind GPT-5.6 Sol and **2.7 points** behind Claude Fable 5.[2] Its measured **$0.94 cost per Intelligence Index task** is 9.4% below Sol and 65.8% below Fable, while Moonshot's **2.8-trillion-parameter**, **1-million-token** model is scheduled to release full weights by **July 27, 2026**.[1][2] K3 is not the overall intelligence champion. It is the open-model champion in waiting, and that distinction is the real story. Moonshot AI released Kimi K3 on July 16, then punctured its own launch hype with a rare admission: the model's overall performance still trails Claude Fable 5 and GPT-5.6 Sol.[1] That does not kill the champion argument. It tells us which crown K3 is actually chasing. Artificial Analysis independently measured K3 at **57.1123**, versus **59.8606** for the routed Fable configuration and **58.8898** for Sol at max reasoning. The same evaluation measured K3 at **62.0 output tokens per second**, **1.99 seconds** to first token, and **$0.9398** per weighted benchmark task.[2][4] Those numbers place K3 inside the frontier, not above it. The real story isn't that Moonshot won every test. It is that a model promised for open-weight release has reached within 3.02% of Sol's independent intelligence score, leads one current independent automation benchmark, offers a 1-million-token context window, and undercuts the premium leaders on measured task cost.[2][5] ### Why This Matters Now The frontier premium depends on scarcity. Kimi K3 challenges that scarcity by combining near-leading intelligence, broad agent performance, million-token context, and a promised weight release in one system. If Moonshot ships usable weights and a workable license, proprietary labs will have to defend not just better scores, but why buyers should accept permanent API dependence. ## The Independent Verdict: Fourth Place Is Not A Defeat Let's be clear: Kimi K3 is not number one on the broadest independent scoreboard available at launch. Artificial Analysis' Intelligence Index v4.1 weights agent tasks at **34%**, coding at **24%**, scientific reasoning at **24%**, and general capability at **18%**. It combines nine evaluations and is text-only and English-only, so it is useful without being universal.[3] K3 ranks fourth because the live class includes Fable 5, Sol at max reasoning, and a second Sol configuration above it. ### Kimi K3's Independent Launch Snapshot Artificial Analysis measurements captured on July 17, 2026. - label: Intelligence Index; value: 57.1; description: #4 of 189 in the current comparison class - label: AutomationBench-AA; value: 52.7%; description: #1 on guardrail-adjusted objective completion - label: Output speed; value: 62.0 t/s; description: Below the 72.7 t/s comparable-model median - label: Time to first token; value: 1.99 s; description: Better than the 2.60 s comparable-model median - label: Cost per AA task; value: $0.94; description: 9.4% below Sol max and 65.8% below Fable with fallback - label: Evaluation output; value: 130M; description: More than twice the 63M comparable-model median The independent category win matters. On AutomationBench-AA, K3 reaches **52.7%** guardrail-adjusted objective completion, ahead of Grok 4.5 at 51.4% and Sol at 51.2%. On the benchmark's separate raw completion measure, Grok leads at 80% and K3 scores 75%, which is why the metric name and methodology cannot be dropped from the claim.[5] ### The Frontier Leaders: Intelligence, Speed And Task Cost - Kimi K3 - Claude Fable 5* - GPT-5.6 Sol 0 - feature: AA Intelligence Index; values: - 57.1 - 59.9 - 58.9 - feature: Output speed; values: - 62.0 t/s - 65.6 t/s - 53.1 t/s - feature: Cost per AA task; values: - $0.94 - $2.75 - $1.04 - feature: Distance from leader; values: - 2.7 points - Leader - 1.0 point *Fable 5 is the max-effort routed configuration with Opus 4.8 fallback. Values are a July 17 live snapshot and can change as providers and measurements update.[2][4]* /images/articles/brand-kit-2026/kimi-k3-open-model-champion/kimi-k3-frontier-gap.webp Three monumental analytical machines stand on a technical frontier, with a crimson-routed challenger just below two taller leaders. Kimi K3 sits within striking distance of the proprietary frontier, but independent scoring still places two leaders marginally ahead. 1664 936 16/9 cover 940 Fourth place looks different when the gap is **2.7 points** and the measured task cost is **65.8% lower**. K3 did not take the intelligence crown. It made the crown look expensive. ## The Radar: Breadth Is K3's Strongest Claim A radar chart is useful only when every axis speaks the same mathematical language. Mixing Intelligence Index points, tokens per second, dollars, Elo, and context length would create a dramatic shape with no defensible meaning. The comparison below uses seven 0-to-100 scores from Moonshot's own launch table instead.[1] ### Moonshot-Reported Capability Profile At Maximum Effort Seven percentage-based benchmarks spanning coding, terminal work, browsing, tool use, automation, science, and vision. Toggle models or switch to the table for exact values. - 0 - 100 - model: Kimi K3; company: Moonshot AI; metrics: program: score: 77.8; unit: %; note: KimiCode harness; terminal: score: 88.3; unit: %; note: KimiCode harness; browse: score: 91.2; unit: %; note: Context compaction at 300K; mcp: score: 84.2; unit: %; note: MCP Atlas; automation: score: 30.8; unit: %; note: Moonshot launch evaluation; gpqa: score: 93.5; unit: %; note: GPQA Diamond; mmmu: score: 81.6; unit: %; note: MMMU-Pro without Python - model: Claude Fable 5*; company: Anthropic; metrics: program: score: 76.8; unit: %; note: Vals AI result; terminal: score: 84.6; unit: %; note: Terminus 2; browse: score: 88; unit: %; note: Published comparator; mcp: score: 84.7; unit: %; note: MCP Atlas; automation: score: 29.1; unit: %; note: Moonshot launch evaluation; gpqa: score: 92.6; unit: %; note: GPQA Diamond; mmmu: score: 81.2; unit: %; note: MMMU-Pro without Python - model: GPT-5.6 Sol; company: OpenAI; metrics: program: score: 77.6; unit: %; note: Vals AI result; terminal: score: 88.8; unit: %; note: Codex harness; browse: score: 90.4; unit: %; note: Published comparator; mcp: score: 83.6; unit: %; note: MCP Atlas; automation: score: 29.7; unit: %; note: Moonshot launch evaluation; gpqa: score: 94.1; unit: %; note: GPQA Diamond; mmmu: score: 83; unit: %; note: MMMU-Pro without Python - key: program; name: Program Bench; description: Repository-level coding - key: terminal; name: Terminal-Bench 2.1; description: Agentic terminal work - key: browse; name: BrowseComp; description: Web research - key: mcp; name: MCP Atlas; description: Multi-tool workflows - key: automation; name: Automation Bench; description: SaaS workflow completion - key: gpqa; name: GPQA Diamond; description: Scientific reasoning - key: mmmu; name: MMMU-Pro; description: Multimodal reasoning *Moonshot-reported launch results. K3 was run at max effort. Several agentic rows use different native harnesses, so the chart compares deployed model-agent systems rather than base models under one standardized harness. Fable includes fallback behavior. The polygon is a capability profile, not an independent composite score.[1][12][13]* The shape tells a more useful story than its area. K3 leads Program Bench, BrowseComp, and Moonshot's Automation Bench row. Fable leads MCP Atlas. Sol leads Terminal-Bench 2.1, GPQA Diamond, and MMMU-Pro. K3 is competitive everywhere, but it does not dominate everywhere. Here is the trap. The unweighted average across these seven selected axes gives K3 **78.2000** and Sol **78.1714**, a meaningless 0.0286-point K3 lead. Add the equally reasonable HLE-Full result and Sol moves ahead, **73.9625** to **73.8625**. A champion claim that flips when one valid axis is added is not a champion claim. It is chart selection. Kimi K3's radar is evidence of breadth. It is not a mathematical crown. LLM Rumors Analysis ## The Architecture: Moonshot Attacked Sequence And Depth K3's **2.8 trillion parameters** are the headline. The architecture is the strategy. Moonshot combines Kimi Delta Attention, or KDA, with Attention Residuals and Stable LatentMoE. The router activates **16 of 896 experts** for a token. Moonshot has not disclosed the model's total active parameter count, so dividing 2.8 trillion by the expert ratio would be fake precision.[1] KDA changes information flow across sequence length. The earlier Kimi Linear research reported up to **75% lower KV-cache use** and as much as **6x decoding throughput** at 1-million-token context in its own comparison. Those are Kimi Linear paper results, not measured K3 production gains.[9] Attention Residuals changes information flow across depth by learning which earlier representations should be retrieved instead of accumulating every layer uniformly.[10] 16 of 896 routed experts activated Moonshot's disclosed Stable LatentMoE routing pattern. Total active parameters remain undisclosed, and the technical report is still pending. /images/articles/brand-kit-2026/kimi-k3-open-model-champion/kimi-k3-kda-attnres-architecture.webp Engraved cutaway of an industrial intelligence stack, with a blank context ribbon crossing layered machinery, vertical retrieval conduits spanning depth, and crimson routes activating a sparse subset of many expert chambers. A conceptual, not-to-scale rendering of Kimi K3's design thesis: Kimi Delta Attention carries long-sequence flow, Attention Residuals retrieve selectively across depth, and sparse routing activates only part of the expert field. 1664 936 16/9 cover 940 Here's the genius: Moonshot attacked the two places where giant models become economically awkward. KDA targets the memory burden across long sequences. Attention Residuals targets the information and communication burden across deep stacks. Stable LatentMoE limits how much of the expert pool is used for a token. Moonshot claims the combined system delivers an approximate **2.5x improvement in overall scaling efficiency** over Kimi K2. It also uses quantization-aware training with MXFP4 weights and MXFP8 activations, and recommends supernodes with **64 or more accelerators** for deployment.[1] Those are first-party disclosures awaiting the promised technical report and weights. The serving layer matters too. Moonshot's Mooncake research separates prefill from decoding and makes KV-cache movement a scheduling problem. The company says K3's official API achieves a cache-hit rate above **90% in coding workloads**.[1][11] Architecture and infrastructure are being designed as one commercial system. ## The Economics: Cheap Per Result, Expensive Per Token K3 is not a bargain-bin model. The official API charges **$0.30 per million cache-hit input tokens**, **$3.00 per million cache-miss input tokens**, and **$15.00 per million output tokens**.[7] Artificial Analysis calls the $3 and $15 list prices somewhat expensive relative to comparable-model medians of $1.75 and $8.40.[2] The real story isn't token price. It is cost per completed high-end task. K3's **$0.94** weighted Intelligence Index task cost sits below Sol's $1.04 and far below Fable's $2.75, even though cheaper and lower-scoring systems beat K3 on pure cost.[2] That advantage comes with a verbosity tax. K3 generated **130 million output tokens** during the Intelligence Index run, versus a 63 million comparable-model median, and the full evaluation cost **$2,690.80**.[2] Max-only reasoning helps capability, but a $15 output rate turns every unnecessary reasoning token into a line item. /images/articles/brand-kit-2026/kimi-k3-open-model-champion/kimi-k3-task-cost-verbosity.webp An engraved compact machine produces a polished mechanism while an enormous crimson-tinted trail of punched paper accumulates behind it beside a balance scale. K3's measured cost per completed task can look competitive even as its unusually verbose outputs create a separate token-volume burden. 1672 941 16/9 cover 940 ### Cost Per Task Is Not Cost Per Accepted Result Artificial Analysis' $0.94 figure is a weighted benchmark cost, not a universal production bill. Teams still need to price retries, tool calls, verification, long reasoning traces, cache-hit rates, and failures. K3's economics look strongest when a workload reuses large prefixes and accepts the model's first completed artifact. The uncomfortable truth is that open models no longer have to be cheap to disrupt the market. They only need to be good enough that control, inspectability, and deployment choice become worth paying for. ## The Open Bet: The Weights Are Still A Promise Moonshot calls K3 the first open 3T-class model and says the full weights will be released by **July 27, 2026**. As of July 17, those weights are not on Moonshot's public Hugging Face model list, no K3 license has been published, and Artificial Analysis still classifies the model as proprietary.[1][2][14] That is why "champion in waiting" is not wordplay. It is a release-state description. /images/articles/brand-kit-2026/kimi-k3-open-model-champion/kimi-k3-open-weights-deadline.webp A monumental black model core waits on rails inside a partly opened industrial gate, with a crimson seal still engaged and empty deployment workshops beyond. K3 is available as a hosted system today, but its open-model status still depends on Moonshot releasing complete weights and a workable license. 1672 941 16/9 cover 940 The hosted product is real today. K3 is available through Kimi.com, Kimi Work, Kimi Code, and the `kimi-k3` API. The API exposes max reasoning at launch, supports text and image input, and documents a default maximum completion of **131,072 tokens** that can be raised within the 1,048,576-token total context limit.[1][6][8] What's often overlooked is the operational coupling. Moonshot warns that multi-turn and tool workflows must return the assistant's earlier thinking and tool state unchanged. Dropping that history or switching models mid-session can destabilize quality.[6] K3 may be open later, but its best behavior could still depend on a carefully matched harness and serving stack. ### What Must Happen Before The Crown Sticks - Moonshot must publish the complete K3 weights by July 27 with a clear license and reproducible inference guidance. - Independent evaluators must rerun K3's strongest coding and workflow claims under matched agent harnesses. - The technical report must disclose active parameters, training details, architecture ablations, and the evidence behind the 2.5x scaling-efficiency claim. - Production teams must test accepted-result cost because max reasoning and high verbosity can erase list-price advantages. - Open-source serving projects must prove that K3 can be deployed outside Moonshot's 64-accelerator supernode recommendation without destroying throughput. ## The Verdict: K3 Targets The Frontier Moat Kimi K3 did not beat Fable 5 or GPT-5.6 Sol on Artificial Analysis. It did not sweep Moonshot's own benchmark table. It is not downloadable today. Those are not footnotes. They define the limits of the claim. They also reveal why the launch matters. K3 is 1.8 index points behind Sol, costs 9.4% less per measured task, runs 16.8% faster in the current provider snapshot, supports a 1-million-token context window, and has a credible independent win on AutomationBench-AA. Against Fable, it gives up 2.7 index points while cutting measured task cost by 65.8%. While competitors protect proprietary margins, Moonshot was building a credible exit ramp from the closed frontier. Axios captured the launch-day industry reaction, but even the excitement came with the correct warning: K3 was only hours old, and early benchmarks could overstate real-world reliability.[15] Call K3 the overall champion and the data says no. Call it open source today and the missing weights say no. Call it the open-model champion in waiting, and every important number points to the same conclusion. The crown belongs to Fable and Sol for now. The moat is what Kimi K3 just put on trial. ## Sources 1. [Kimi K3: Open Frontier Intelligence](https://www.kimi.com/blog/kimi-k3) 2. [Kimi K3: Intelligence, Performance & Price Analysis](https://artificialanalysis.ai/models/kimi-k3) 3. [Artificial Analysis Intelligence Benchmarking Methodology](https://artificialanalysis.ai/methodology/intelligence-benchmarking) 4. [Kimi K3 Provider Benchmarks](https://artificialanalysis.ai/models/kimi-k3/providers) 5. [AutomationBench-AA: Agentic SaaS Workflow Benchmark](https://artificialanalysis.ai/evaluations/automationbench-aa) 6. [Kimi K3 API Quickstart](https://platform.kimi.ai/docs/guide/kimi-k3-quickstart) 7. [Kimi K3 API Pricing](https://platform.kimi.ai/docs/pricing/chat-k3) 8. [Model Configuration: Kimi Code](https://www.kimi.com/code/docs/en/kimi-code/models.html) 9. [Kimi Linear: An Expressive, Efficient Attention Architecture](https://arxiv.org/abs/2510.26692) 10. [Attention Residuals](https://arxiv.org/abs/2603.15031) 11. [Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving](https://arxiv.org/abs/2407.00079) 12. [DeepSWE Leaderboard](https://deepswe.datacurve.ai/) 13. [Program Bench](https://www.vals.ai/benchmarks/programbench) 14. [Moonshot AI Models](https://huggingface.co/moonshotai/models) 15. [China's Open-Weight Kimi Model Stuns AI World With Frontier-Level Results](https://www.axios.com/2026/07/16/moonshot-kimi-ai-china-model-openai-anthropic) *Last updated: July 17, 2026* --- *Source: [LLM Rumors](https://www.llmrumors.com/news/kimi-k3-open-model-champion-in-waiting)*