The ORCFLO Indexis an independent benchmark that evaluates large language models the way business professionals actually use them — across real-world tasks spanning analysis, writing, extraction, summarization, and behavioral reliability. Each model is scored on three dimensions (quality, cost, and speed) by a panel of four independent judges. This report evaluates Gemini 3 Pro (Preview) in the context of 32 models from ORCFLO Index — May 10, 2026 cohort · 32 total models tested · Anthropic, Google, OpenAI, Mistral.
Gemini 3 Pro (Preview)
Gemini 3 Pro (Preview) lands at #2 of 32 with a quality score of 93.17, trailing only GPT 5 (95.04) and edging out GPT 5.5 and Claude Opus 4.6. It leads the field outright in Extraction (#1) and posts leader-tier results across six of eight categories. Cost is moderate at $0.0081 per case (#18), but response time of 21.9 seconds (#28) is a notable drag. A near-top-of-field quality profile at a fraction of GPT 5's cost, provided latency is acceptable.
Key Findings
- Second-highest overall quality (93.17) in a 32-model field, with GPT 5 the only model scoring higher (95.04) — and at roughly 3.3x the cost per case.
- Outright category leader in Extraction (96.5, #1) and leader-tier in Analysis (#4), Summarization (#5), Writing (#6), Hallucination (#5), and Refusal Calibration (#7).
- Cost-to-quality ratio is unusually strong: at $0.0081 per case (#18), it costs less than GPT 5.5 ($0.0419), Claude Opus 4.6 ($0.0221), and GPT 5 ($0.0267) while outranking all but GPT 5 on quality.
- Latency is the principal weakness: 21.9 seconds per response ranks #28 of 32, slower than peer GPT 5.1 (7.6s) and most mid-tier alternatives.
Model Performance: Quality & Cost
The chart below plots quality against cost for all 32 models in the ORCFLO Index. Each dot represents the average quality score a model achieved across the full basket of real-world business tasks, alongside the cost in credits to complete the entire test suite. Models in the upper-left quadrant deliver the highest quality at the lowest cost.
Model Performance: Quality & Time Elapsed
Quality alone doesn’t tell the full story — response time determines whether a model is viable for time-sensitive workflows. The chart below plots each model’s quality score against the total time required to complete the test suite. Models in the upper-left deliver the best quality with the least delay.
Category Scorecard
The ORCFLO Indexevaluates models using real-world business tasks — not academic puzzles or synthetic benchmarks. Each test case is designed to expose specific differences in how models handle the work professionals actually do. Scores are averaged across each category and ranked independently across all 32 models.
| Category | Score | Rank | Tier |
|---|---|---|---|
| Abilities — Core language tasks: what the model can produce when given a well-formed prompt. | |||
| AnalysisReasoning, strategic judgment, disqualifying-factor detection | #4 | Leader | |
| ExtractionField accuracy, null handling, format compliance, zero fabrication | #1 | Leader | |
| SummarizationCompression quality, key-point retention, length compliance | #5 | Leader | |
| WritingTone, structure, persuasion, audience adaptation | #6 | Leader | |
| Behaviors — How the model acts under pressure: reliability, compliance, and restraint. | |||
| HallucinationFabrication detection, factual grounding, source fidelity | #5 | Leader | |
| Instruction FollowingConstraint adherence, format compliance, multi-part directives | #9 | Strong | |
| Refusal CalibrationAppropriate refusal vs. over-refusal on legitimate requests | #7 | Leader | |
| Stability — Repeatability and predictability across identical inputs. | |||
| Output ConsistencyRun-to-run reproducibility, format stability, score variance | #21 | Contender | |
Strengths and Cautions
Strengths
- Best-in-field Extraction performance (96.5, #1 of 32) makes it the top choice for structured data pulls, form parsing, and document field capture.
- Low hallucination rate (95.2, #5) paired with strong Refusal Calibration (91.7, #7) supports use in regulated or high-stakes content workflows where factual reliability matters.
- Quality-per-dollar advantage among top-tier models: matches or beats the quality of models costing 3-5x more, including GPT 5.5 ($0.0419) and Claude Opus 4.6 ($0.0221).
Cautions
- Speed rank of #28 (21.9s average) rules it out for latency-sensitive use cases; GPT 5.1 delivers comparable quality (#5) in 7.6 seconds.
- Output Consistency falls to contender tier (87.2, #21), the only category where it drops out of the top ten — a concern for workflows requiring deterministic, repeatable outputs.
- Instruction Following (93.1, #9) is the weakest of its leader-tier abilities, trailing several peers on strict prompt adherence.
Head-to-Head: Frontier Models
Gemini 3 Pro (Preview) is Google’s near-top performer in the ORCFLO Index. The table below compares it against the top-performing models from each major provider. Tier assignments use 25% quartiles across the full 32-model field.
| Model | Quality Avg | Quality Rank | Cost Rank | Speed Rank |
|---|---|---|---|---|
| GPT 5 | 95.0 | #1 | #31 | #32 |
| Gemini 3 Pro (Preview) | 93.2 | #2 | #18 | #28 |
| GPT 5.5 | 93.0 | #3 | #32 | #29 |
| Claude Opus 4.6 | 92.8 | #4 | #29 | #26 |
| GPT 5.1 | 92.7 | #5 | #22 | #13 |
| Gemini 2.5 Pro | 92.0 | #6 | #19 | #27 |
| GPT 5.2 | 91.9 | #7 | #23 | #19 |
When to Use Gemini 3 Pro (Preview)
The ORCFLO Index
This evaluation covers 40 cases across 8 categories. All tasks are text-only and English-only. Code generation, multi-turn conversation, multimodal tasks, and agentic workflows are not tested. Each contestant is scored by a panel of four independent judges — Gemini 2.5 Pro, Claude Opus 4.7, GPT 5.5, and Mistral Large — with final scores averaged across all four. Cost and speed measurements reflect API pricing and latency as of the test date (May 10, 2026) and will change as providers update their offerings.
How We Test
The ORCFLO Indexevaluates large language models across three independent dimensions — quality, cost, and speed — using real-world business tasks designed to expose the differences that matter for model selection. Each model is scored by a panel of four independent judges to reduce single-model bias.
- Test Cases
- 40 cases across 8 categories spanning Abilities (Analysis, Extraction, Summarization, Writing), Behaviors (Hallucination, Instruction Following, Refusal Calibration), and Stability (Output Consistency).
- Judge Panel
- Gemini 2.5 Pro, Claude Opus 4.7, GPT 5.5, and Mistral Large. Each judge scores independently. Final score is the average across all four.
- Scoring
- Three independent ranks: quality (higher is better), cost (lower is better), speed (faster is better). No composite score — composites hide the tradeoffs that drive model-selection decisions.
- Tier Definitions
- LeaderQuality ≥ 90.8Ranks 1–8Strong≥ 85.9Ranks 9–16Contender≥ 80.9Ranks 17–24Trailing< 80.9Ranks 25–32