Today's price · per 1M tokens
Input
$0.950
per 1M tokens
Output
$4.00
per 1M tokens
Blended
$1.71
blended $/1M — 3:1 weighted input:output
Cached input
$0.160
17% of input — prompt caching
Price receipt
We read Fireworks's own pricing page on and found Kimi K2.6 listed with a price on the same row — that read is where the number above comes from. We keep the page text we read; its content hash is 4f5353269c.
Overview
Kimi K2.6 on Fireworks. $0.95/$4.00 per 1M; cached $0.16.
Deploy this open model on rented GPUs
Open weights mean you can self-host instead of paying per-token API prices. These platforms let you serve it on demand — often cheaper at scale.
Partner links — we may earn a commission if you sign up, at no cost to you. We only list platforms we'd recommend regardless.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterSources: Vellum LLM Leaderboard (GPQA Diamond, BrowseComp, OSWorld-Verified, Humanity's Last Exam, AIME 2026) · Moonshot model card (SWE-Bench Verified, ARC-AGI 2, AIME 2025, MMMLU, BFCL, MATH 500)
- GDPval-AA v2: 1046.4 ELO (Real-world work tasks, human baseline = 1000)
- Epoch Capabilities Index: 151.6 (CI 149.9–153.4) (Epoch AI's index fitted across many benchmarks)
- FrontierMath Tiers 1–3: 57.2% (Research-level maths (funded by OpenAI, run by Epoch); best run across the effort settings Epoch tested)
- FrontierMath Tier 4: 25.6% (Hardest research maths; best run across the effort settings Epoch tested)
- GDP.pdf: 13% (reasoning effort, measured 2026-11-22) — Tasks over long professional documents; share of tasks that pass every criterion
- AutomationBench-AA: 13% (reasoning effort, measured 2026-11-22) — Multi-step work across simulated business apps, without breaking the rules set
- MLCR-AA: 11.1% (reasoning effort, measured 2026-11-22) — Reasoning over long, fragmented medical records; pass rate
- Humanity's Last Exam: 37.5% (reasoning effort, measured 2026-11-22)
- GPQA Diamond: 91.1% (reasoning effort, measured 2026-11-22)
- AA-LCR (long-context reasoning): 81% (reasoning effort, measured 2026-11-22)
- CritPt (research-level physics): 8% (reasoning effort, measured 2026-11-22)
- IFBench (instruction following): 76% (reasoning effort, measured 2026-11-22)
- τ²-Bench Telecom: 95.9% (reasoning effort, measured 2026-11-22)
- Terminal-Bench Hard: 43.9% (reasoning effort, measured 2026-11-22)
- Terminal-Bench 2.1: 65.9% (reasoning effort, measured 2026-11-22)
- Terminal-Bench 4.0: 0.5% (reasoning effort, measured 2026-11-22)
- SciCode: 51.5% (reasoning effort, measured 2026-11-22)
- MMMU-Pro: 79.4% (reasoning effort, measured 2026-11-22)
- Ranks #34 of 76 benchmarked models by average score
- Ranks #88 of 207 comparably-measured models by percentile score, across 5 independent measurements
- Ranks #107 of 207 comparably-tested models by normalized performance per dollar
- Strongest at Humanity's Last Exam — 54%, #16 of 81
90 tokens/sec output
Throughput: Artificial Analysis via Vellum, recorded 2026-08-01 · Source: Artificial Analysis (artificialanalysis.ai)
Source: Artificial Analysis, Vellum LLM Leaderboard, Moonshot model card · updated · See full rankings →
Specifications
- Provider
- Fireworks
- Context window
- 256K tokens
- Modality
- text, image, video
- Parameters
- Proprietary
- Open source
- Yes — open weights available
- Released
- Status
- Current
- Last updated
- Tags
Availability verified: — listed on Fireworks's own page