Grok 4.6 vs Kimi K3

head-to-head

what do the benchmarks actually say?

On SWE-bench Verified — real, human-validated GitHub issues resolved end-to-end — Grok 4.6 posts 95.6% against 93.4% for Kimi K3, a 2.2-point gap. Verified is the closest public proxy for "can it fix a real bug in a real repo without help", which is why it anchors our ranking.

A few points either way is real but not decisive: within that band, the agent scaffolding around the model — how it retrieves files, runs tests, and retries — often matters as much as the base model. Treat the gap as a lean, not a verdict.

which is cheaper to run?

Grok 4.6 is the cheaper model: $2 per 1M input tokens ($6 output) versus $3 ($15 output) for Kimi K3 — roughly 1.5× less on input. Coding workloads are output-heavy — agents write diffs, tests and retries — so weight the output rate more than the input rate when you estimate a monthly bill.

Kimi K3 ships open weights. Beyond the per-token discount, that means you can self-host it, fine-tune it, pin a snapshot so the model can't change under you, and keep code on your own infrastructure — none of which is possible with Grok 4.6, which is API-only. For teams with compliance constraints that difference outweighs any benchmark point.

when to pick each

how were these scores verified?

We only print a number once it's confirmed against a primary source or an independent evaluation, and each row on our leaderboard records which kind it is:

• Grok 4.6: Independent (vals.ai, benchmark updated 2026-08-12, mini-swe-agent bash-only harness): SWE-bench Verified 95.60% ±0.92. Entered ranked 2026-08-13, one day after launch, and it enters near the top: 4th of the 82 systems vals.ai has run, behind Claude Opus 5 (97.00% ±0.76), DeepSeek V4 Pro 0813 (96.40% ±0.83) and GPT-5.6 Sol (96.20% ±0.86). Read the gap to Claude Fable 5 below it (95.00% ±0.98) as a tie: 0.6 points against a pooled SEM of about 1.34 is well inside the margin of error, and we rank Grok 4.6 higher only because it scored higher. Same against Sol above it, a 0.6-point gap. This is the second time running that xAI shipped a Grok with no SWE-bench Verified number of its own and vals.ai supplied one: Grok 4.5 waited nine days, Grok 4.6 waited one. SpaceXAI still publishes no SWE-bench Verified or SWE-bench Pro figure, so the ranked score here is entirely vals.ai's measurement with no vendor claim to disclose against. What xAI did publish, all vendor-reported on its own table: AA Intelligence Index 61, CursorBench v3.2 69.9%, DeepSWE v1.1 65.9%, FrontierCode v1.1 Extended 61.3%, APEX-Agents 57.5%, APEX-SWE 56.4%, Terminal-Bench v3.0 26%, GDPVal-AA v2 1753, AA-Briefcase 1577, Harvey LAB (Vals) 15.8%. That table's shape still holds and is worth reading against this score: Grok 4.6 takes the best result on the three knowledge-work evals and loses the two hardest agentic-coding ones by wide margins (DeepSWE 65.9% vs Sol's 73%, Terminal-Bench v3.0 26% vs Sol's 34.6%). Terminal-Bench v3.0 is NOT comparable to the 83.3% TB2.1 figure on the Grok 4.5 row: on v3.0 Grok 4.5 scores 15.7% and the whole field tops out near 34%. On cost it is the standout of the top five, $0.78 per test against $1.29 for Opus 5, $1.15 for Sol and $2.05 for Fable 5, though it is also the slowest of them at 604s. Pricing unchanged from Grok 4.5 at $2/$6 per 1M. See /p/grok-4-6-frontier-benchmarks-terminal-bench-gap/.
• Kimi K3: Independent (vals.ai, mini-swe-agent bash-only harness): SWE-bench Verified 93.40% ±1.11. Verified Jul 18, 2026 — vals.ai had not run K3 at our Jul 17 check and now ranks it third overall (above GPT-5.6 Luna at 93.0% and Claude Opus 4.8 at 88.6%), making it the highest-scoring open-weight coder we track. Moonshot published no SWE-bench Verified score of its own, only Terminal-Bench 2.1 88.3% on its KimiCode harness, so we rank on the independent number per our standing rule. Corroborated by a second independent evaluator: Artificial Analysis, which runs its own tests, has K3 4th on GDPval-AA v2 (Elo 1684), 2nd on AA-Briefcase (Elo 1545) and 4th of 187 on its Intelligence Index v4.1 (57.1). It also classifies K3 as proprietary outright. AA measured Terminal-Bench 2.1 at 85.02% against Moonshot's own claim of 88.3%, a 3.3-point vendor overstatement, which is why we print the independent figure. Blind developer voting on Arena WebDev ranks it 1st of 99. Price confirmed on Moonshot's own pricing page: $3/$15 per 1M, cache-hit input $0.30 (platform.kimi.ai, checked Jul 19, 2026). **open: false, corrected Jul 19, 2026** — we had this flagged as open weights, which put it top of our "best open-source coding model" page as a model nobody can download. Checked Jul 19: no K3 repo under huggingface.co/moonshotai (18 repos, none K3) and github.com/MoonshotAI/Kimi-K3 returns 404. Moonshot's own blog still says "The full model weights will be released by July 27, 2026." Until they actually ship, open is a roadmap item, not a fact, and this row is an API model. Flip back to open: true when the weights land. Re-checked Jul 27, 2026 (Moonshot's own deadline day): huggingface.co/moonshotai still lists no K3 repo, newest is Kimi-K2.7-Code. Score unchanged at 93.40% on vals.ai. **open: true, restored Jul 27, 2026 13:31 UTC**: the weights actually landed at huggingface.co/moonshotai/Kimi-K3 later the same day, 96 safetensors shards, 1,560,998,983,759 bytes (about 1,454 GiB), under a bespoke "Kimi K3 License" that is permissive but not Apache 2.0 (separate agreement required above $20M MaaS revenue; attribution required above 100M MAU). Verified against the Hugging Face API and config.json, not a press report. One caveat that belongs on the record: the release is 4-bit MXFP4 only for the routed experts (quantization-aware trained from the SFT stage, format mxfp4-pack-quantized, group size 32), with attention, shared experts, LM head and vision tower left at bf16, and there is no base checkpoint, so this is open for inference rather than open for training. Score unchanged at 93.40%; nothing was re-ranked by this flag. See /p/kimi-k3-weights-landed-4-bit-only/.

Full reviewsGrok 4.6, decodedKimi K3, decoded

Ranked on our AI Coding Leaderboard, updated 2026-08-13. Scores are confirmed against primary sources; prices are per 1M input tokens and can change.