Same as 1.8 Flash — whole workspaces and extended agent threads without a separate context tier.
Model release
Boongle 1.8
Pro
Our highest-intelligence first-party model for Boongle IDE — built for long-running agents, deep refactors, and instructions that would trip 1.8 Flash. Same 1M context window, stronger reasoning per step, modest BoongleBench gains — still below frontier API models like Claude Fable 5.1.
1Mcontext window
Same window as 1.8 Flash — large repos and long agent threads, with Pro tuned for quality over raw turn speed.
79.6BoongleBench composite
+1.2 vs 1.8 Flash on our internal harness. Claude Fable 5.1 leads at 83.6 composite on the same leaderboard.
Official release
Shipped 18 September.
Boongle 1.8 Pro is live in Boongle IDE as of 18 September 2026. It sits above 1.8 Flash in the first-party lineup: same Boongle agent harness, but trained and tuned for sustained intelligence rather than maximum turn speed.
Short announcement: /announcements/boongle-1.8-pro. This page is the full technical story — specs, benchmarks, and when to pick Pro over Flash.
Specifications
Pro tier, defined.
Still responsive in the IDE tool loop; Pro trades a little latency for much stronger reasoning per step.
Per 1M input / output tokens on the Boongle AI API. Included usage on Pro+ and Ultra IDE plans.
Boongle IDE: select Boongle 1.8 Pro in chat, inline edit, or Agent when your plan includes it. Boongle AI API: same model at $3.00/M input and $12.00/M output for programmatic access.
Vs 1.8 Flash
When Pro wins.
Flash remains the default for fast daily edits and high-volume agent turns. Pro is for work where quality and instruction-following matter more than raw speed.
Metric
1.8 Flash
1.8 Pro
Context window
1M
1M
SWE-Bench Multilingual
78.6
80.8
Terminal-Bench 2.1
71.5
74.6
Agent bench v3.1
61.9
64.8
Speed index
94.0
86.0
Cost efficiency
88.0
72.0
BoongleBench composite
78.4
79.6
API (in / out per 1M)
$0.15 / $0.30
$3 / $12
Frontier reference: on the same BoongleBench harness, Claude Fable 5.1 scores 83.6 composite (SWE 82.4, Terminal 86.2, Agent 72.5). Pro is a first-party upgrade over Flash — not a replacement for top-tier API models. See the full leaderboard.
BoongleBench
Pro scores.
Internal BoongleBench runs on the March 2026 harness — same methodology as our public leaderboard, with Pro evaluated in the full Boongle IDE agent stack.
Capabilities
What Pro is for.
- Long-running agents — multi-step plans that stay coherent across dozens of tool calls.
- Complex instructions — nested requirements, style guides, and “do not touch X” constraints held reliably.
- Large refactors — cross-package changes with consistent architecture decisions.
- Deep debugging — hypothesis → test → fix loops on gnarly production issues.
- 1M context — same window as Flash, with Pro prioritizing depth over turn latency.
Get started
Try Pro today.
Boongle 1.8 Pro is available in Boongle IDE for Pro+ and Ultra subscribers. Open the model picker and choose Boongle 1.8 Pro, or read the shipped announcement for pricing and benchmark highlights.