AI Evaluation
Kimi K3 Enters Artificial Analysis's Coding Agent Index at 57
Artificial Analysis added Kimi K3 in Kimi Code CLI to its Coding Agent Index on July 18 with a composite score of 57 and a reported $3.18 average API cost per task. The result is independent of Moonshot's launch claims but depends on Artificial Analysis's three-benchmark composite and harness configuration.
What happened and why it matters
This follow-up replaces launch-day self-reporting with an outside test, without turning one composite into a verdict on the whole model.
Primary source
Primary reference: Artificial Analysis: Coding Agent Index. Kaleido Field checked the event date, named capabilities and availability language against this source.
| Source date | Kimi K3 result reported July 18, 2026 |
|---|---|
| Checked by Kaleido Field | July 19, 2026, 09:18 CST |
| What this source supports | independent benchmark operator result with harness and cost boundaries for Kimi K3 Coding Agent Index 57 3.18 per task Artificial Analysis |
| What it does not prove | It does not prove a universal product ranking, full regional availability, or performance on every visual intelligence task. |
The new result
Artificial Analysis reports Kimi K3 running through Kimi Code CLI at 57 on its Coding Agent Index. The July 18 result was described as joint fifth at publication time and the strongest tested open-weight configuration in that table.
The operator also reports an average pay-per-token API cost of $3.18 per task. That estimate uses observed token consumption and provider pricing; it is not a subscription price or a full measure of engineering cost.
How the index is built
The composite averages pass@1 across DeepSWE, Terminal-Bench v2 and SWE-Atlas-QnA. Artificial Analysis says model and harness both matter, which is why the result should be cited as Kimi K3 in Kimi Code CLI rather than a context-free model score.
Cache hits, provider routing, token mix and agent runtime can change effective cost. The site explicitly notes that its cost chart omits infrastructure, engineering and supervision.
Why this is separate from launch coverage
Kaleido Field covered Moonshot's Kimi K3 release on July 18 using company evidence. This article records a later independent benchmark entry and targets a narrower question: how one coding-agent configuration performed under an outside test.
The result does not validate Kimi K3's visual understanding, one-million-token context claims or general research ability. Leaderboard positions can also move as new models and reruns enter the index.
Evidence boundary
This page reports a dated event from a named primary source. Company specifications and adoption statements remain attributed claims unless independent evidence is cited above.
FAQ
What is the practical answer?
Artificial Analysis added Kimi K3 in Kimi Code CLI to its Coding Agent Index on July 18 with a composite score of 57 and a reported $3.18 average API cost per task. The result is independent of Moonshot's launch claims but depends on Artificial Analysis's three-benchmark composite and harness configuration.
What source does this article use?
The primary source is Artificial Analysis: Coding Agent Index. Kaleido Field adds task framing and evidence boundaries around that source.
Where should the user verify the answer?
Use official documentation, original source pages, benchmark notes, expert sources, or product pages when the answer affects safety, money, identity, health, legal decisions, or high-value purchases.