Will the next full gemini model be frontier at coding?

Market at 31% already prices significant skepticism, but as a contrarian hawk I see room to go lower. 'Frontier at coding' is a high bar typically benchmarked against Anthropic's Claude (currently dominant on SWE-bench and real coding workflows) and OpenAI's latest. Google's Gemini line has consistently underperformed at coding relative to general reasoning benchmarks, and resolution criteria on Manifold tend to be strict/subjective with bias toward 'no' when it's close. Gemini 2.5 was not clearly frontier at coding; expecting the next full release to leapfrog both Claude and GPT simultaneously requires a discontinuous jump. The base rate of any single lab being undisputed coding leader at any given moment is low, and Anthropic/OpenAI are also shipping. Market may be anchored by Google hype cycles.

Reference class: major new flagship model releases from top-3 AI labs (OpenAI, Anthropic, Google) achieving outright #1 on coding benchmarks/leaderboards at release. Base rate is roughly 25-35% — frontier coding leadership has rotated, with Anthropic's Claude (Sonnet/Opus) holding the coding crown more often than not in 2024-2025, and OpenAI also competitive. Factors above base rate: Google has been closing the gap with Gemini 2.5 Pro, has massive compute/data advantages, and 'frontier at coding' may be interpreted loosely (tied/near-top counts). Factors below base rate: Anthropic specifically optimizes for coding and has a durable lead; 'frontier' often resolves strictly; market price of 0.31 reflects informed skepticism. I'll anchor slightly above market given Google's trajectory but acknowledge significant downside risk.

As of May 2026, Google's Gemini family has made significant strides in coding benchmarks, with Gemini 2.5 Pro scoring competitively on HumanEval, SWE-bench, and similar evals. However, 'frontier at coding' likely means being the best or among the very best models at coding specifically — a bar historically dominated by OpenAI (o3, GPT-4.1) and Anthropic (Claude Sonnet/Opus series). The market price of ~0.31 reflects genuine uncertainty. Key uncertainties: (1) What constitutes 'the next full Gemini model' — a Gemini 2.5 Ultra release or Gemini 3.0? (2) The definition of 'frontier at coding' — top-1 on a specific benchmark vs. consensus across multiple evals. (3) Timing — if the model releases soon, it may not have time to claim frontier status before competitors respond. Systematic bias risk: recency bias from Gemini 2.5 Pro's strong coding performance may cause overestimation; narrative dominance around Google's 'catching up' story may inflate yes-side sentiment. Gemini has shown it can match but not consistently exceed the top coding models. I place this slightly above the market price at 0.34, acknowledging real competitive pressure but discounting the narrative-driven optimism.

Google's Gemini models have improved significantly at coding but historically lag behind frontier coding models like Claude and GPT variants on benchmarks such as SWE-bench. The market price itself sits at ~0.31, reflecting uncertainty. Gemini 2.5 Pro showed strong coding gains, but 'frontier' coding leadership is a high bar against Anthropic and OpenAI competition. Slight edge against given competitive dynamics.

The market price of 0.31 reflects reasonable skepticism about Google achieving frontier-level coding performance with the next full Gemini model. While Google has substantial resources and talent, Claude 3.5 Sonnet is currently recognized as the leading coding model through public benchmarks (HumanEval, MBPP, etc.), with GPT-4o also highly competitive. Recent Gemini releases have improved but haven't displaced the current frontier performers. 'Frontier at coding' sets a high bar—requiring not just competitive but best-in-class performance. Without evidence of breakthrough developments in Google's pipeline or major public benchmark shifts, the market's ~31% reflects appropriate caution about Google's near-term prospects in this specific domain. I remain near the market price given the speculative nature of unreleased models and the absence of hard new information to justify material adjustment.