[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]

AI use is no longer a competitive advantage in commercial real estate. It is the floor. That's the takeaway from Adventures in CRE's Summer 2026 assessment, which says AI use is now "the expected baseline in a growing number of firms" across acquisitions, development, management, investor relations, and brokerage. For managers, the question has shifted from whether teams use AI to how reflexively they do so.

The model tier that matters for CRE operators

For CRE teams choosing large language models, Adventures in CRE's Summer 2026 tracker ranks options using the Artificial Analysis Intelligence Index, a composite of benchmarks including MMLU-Pro, GPQA Diamond, MATH-500, and long-context reasoning tests. Claude Opus 5, running in Adaptive Reasoning at Max Effort, tops the July 2026 leaderboard with a score of 60.7, priced at $5.00 per million input tokens and $25.00 per million output tokens. OpenAI's GPT-5.6 Sol in max mode ranks fourth at 58.9, with output pricing at $30.00 per million tokens. Google's Gemini 3.5 Flash leads on throughput at 222 tokens per second, at $1.50 input and $9.00 output per million tokens.

That spread matters for document-heavy workflows. Lease abstraction, financial model generation, and offering memorandum drafting all depend on long-context reasoning and structured output. Adventures in CRE weights intelligence over speed in its composite scoring, because accuracy carries higher stakes than latency in deal underwriting.

From workflow efficiency to market recovery

Jack Mullen, writing for Forbes Finance Council in June 2026, argues that AI could serve as a catalyst for a broader CRE rebound, with faster underwriting cycles, better asset-level analytics, and more efficient capital deployment reducing friction at precisely the points where deals have historically stalled. That framing repositions AI spend from a cost-center question to a deal-velocity question. A team that closes underwriting in days rather than weeks because its analysts run AI-assisted models is not just more efficient; it is more competitive for assets in thin-inventory markets.

The firms most likely to benefit from any market rebound are those already embedding AI into core transaction workflows, not those waiting for the tools to mature further, according to Forbes Finance Council. Adventures in CRE also notes that AI lets non-technical professionals build bespoke analytical tools quickly and at low cost - a meaningful shift for mid-market operators who cannot staff full data-science teams but face the same modeling demands as institutional players.

What the tooling market looks like right now

Beyond model selection, Adventures in CRE categorizes tools across vision, reasoning, tool-use, long-context handling, and whether the model is open-weight. Open-weight models matter because they can be deployed within a firm's own infrastructure, keeping rent rolls and financial projections off third-party servers.

The tracker updates at least quarterly, and the pace shows why. Claude Opus 5 and GPT-5.6 both carry July 2026 release dates, meaning the top of the capability curve moved within the past month alone. For technology buyers supporting CRE firms, vendor lock-in risk is now a real evaluation criterion.

Spencer Burton, co-founder of CRE Agents and co-creator of the AI.Edge training community cited in the piece, said the firm maintains the tracker as a service rather than an endorsement. That appetite for neutral guidance on tool selection is itself a measure of how far adoption has progressed.

Why this matters for managers

Audit current workflows against the capability dimensions that matter most: long-context reasoning for lease and document work, tool-use for financial model integration, and vision for site and property analysis. Match the model to the actual task, not just the benchmark rank.

Price total inference cost at scale before committing. Claude Opus 5 Max Effort and GPT-5.6 Sol differ on output cost ($25 vs. $30 per million tokens); at high query volumes, that gap compounds.

Evaluate open-weight deployment if deal data sensitivity is a governance concern. Third-party API calls mean data leaves your environment; on-premise or private-cloud deployment changes that calculus.

Treat AI tool selection as a recurring procurement decision, not a one-time buy. With top-tier models releasing on a monthly cadence as of mid-2026, the evaluation cycle needs to match the market's pace.