Breaking

Moonshot's Next Kimi K3 Model Expected to Rival Anthropic's Claude Opus 4.8

July 16, 2026 at 10:31 EDT

Foundation Models
Open Source
AI Agents

Moonshot AI's forthcoming Kimi K3 is expected to close the gap with Anthropic's Claude Opus 4.8, positioning what could become China's largest open-weight model as a direct challenger to the leading closed systems from U.S. labs. According to reporting drawing on the Financial Times, the model is anticipated to match or exceed Opus 4.8 on key tasks and is due for release within days.

Frontier AI · Open vs Closed

Kimi K3 aims to close the gap with Claude Opus 4.8

Moonshot AI's forthcoming open-weight model — reportedly in the two-to-three-trillion-parameter range — is expected to rival or exceed one of the leading closed U.S. systems. Release is said to be imminent, but no official benchmarks have been published.

2.8T

parameters cited for K3 (leaked)

token context window (leaked)

$31.5B

valuation Moonshot is reported to target

Parameter scale: K2 → K3

Total parameters, in trillions. K3 is expected to push roughly 3× beyond the K2 series.

~1T

K2 series
(mid-2025)

~2.8T

K3
(expected)

The matchup

Kimi K3 · expected

Params: ~2–3T (2.8T cited)
Context: 1M tokens
Access: Open-weight
Status: Pre-release

Claude Opus 4.8 · released

Params: Not disclosed
Pricing: $5 in / $25 out per M tokens
Access: API, closed
Status: Released May 28, 2026

Why it matters

K2 developers report handling many tasks otherwise routed to Opus at a fraction of the cost, with stable long multi-step workflows and self-hosting flexibility.

The caveat

No official weights, benchmarks or pricing have been released. Claims rest on leaks and app-store descriptions, and independent verification remains outstanding.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.