Claude Sonnet 4.6, the top-tier Sonnet model Anthropic released on February 17, 2026, is drawing developer attention after appearing relatively low on some LLM leaderboards following the arrival of its successor, Sonnet 5.
February 17, 2026 · Anthropic
Claude Sonnet 4.6: near-Opus power at Sonnet prices
Anthropic's fast, cost-efficient middle-tier model gains ground in coding, computer use, and agentic work — even as its own successor, Sonnet 5, later pushes it down relative-scoring leaderboards.
$3 / $15
Input / output per million tokens — held flat from Sonnet 4.5
1M
Token context window (beta, API-centered)
70%
Of developers prefer 4.6 over 4.5; 59% over Opus 4.5
Computer use: the standout leap
OSWorld-Verified — Sonnet 4.6 nearly matches Opus 4.6, far ahead of its predecessor
SWE-bench Verified — coding
4.6 closes almost all the gap to Opus 4.6
Where it shines
Spreadsheet operations & web-form automation
Financial analysis & top-ranked office tasks (GDPval-AA)
Iterative work on complex codebases
Competitive with models like GPT-5.2
Where it lags
Creative writing "flatter and more generic"
Can be overly argumentative
Occasional basic-logic failures
1M context limited on the web version
Why 4.6 now sits low on leaderboards
Sonnet 5's arrival around June–July 2026 surpassed 4.6 on some tasks, causing 4.6 to slip on arena-style boards that use relative scoring .
Yet 4.6 is still widely chosen for its balance of price and speed — and observers caution that a leaderboard position does not equal practical usefulness.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…