TIILanguage model
Falcon 180B
Falcon 180B has no overall index value yet because there are too few independent benchmark results.
PopularityRank 633 of 785
25out of 100same as last month
based on search, usage and attention
Summary
Falcon 180B remains a raw model for further fine-tuning, not a finished chat assistant. The dense decoder released by the Technology Innovation Institute in 2023 was trained on 3.5 trillion tokens from RefinedWeb and curated sources but has only a 2K context.
The official model card estimates about 400GB of memory for fast inference and recommends additional adaptation plus application safeguards. The Falcon-180B TII license contains its own acceptable-use rules.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 42 published benchmark results
Page 1 of 4
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
Hugging Face Open LLM Leaderboard aggregate (Show measurement, test conditions, and source)
- Source value
- 67.85
- Score
- 67.85
- Metric
- percent
- Unit
- %
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- Hugging Face Falcon 180B evaluation retrospective
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Official benchmark leaderboard. Falcon 180B base pretrained model; Open LLM Leaderboard composite after the leaderboard added two benchmarks in November 2023, as reported in the retrospective. Component scores are not transcribed here.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 85.4
- Score
- 85.4
- Metric
- Con
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 3 reports the general-domain task. For the ICL values, the prompt includes one human-written sample; Aver is the mean across the four dimensions.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 95.7
- Score
- 95.7
- Metric
- Rel
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 3 reports the general-domain task. For the ICL values, the prompt includes one human-written sample; Aver is the mean across the four dimensions.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 34.8
- Score
- 34.8
- Metric
- Cov
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 3 reports the general-domain task. For the ICL values, the prompt includes one human-written sample; Aver is the mean across the four dimensions.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 55.1
- Score
- 55.1
- Metric
- Rep
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 3 reports the general-domain task. For the ICL values, the prompt includes one human-written sample; Aver is the mean across the four dimensions.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 67.8
- Score
- 67.8
- Metric
- Aver
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 3 reports the general-domain task. For the ICL values, the prompt includes one human-written sample; Aver is the mean across the four dimensions.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 57.2
- Score
- 57.2
- Metric
- 0-shot
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. General-domain question generation, 0-shot setting. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy general-domain question generation (Show measurement, test conditions, and source)
- Source value
- 18.4
- Score
- 18.4
- Metric
- ↓(%)
- Unit
- %
- Benchmark version
- ACL 2024, Table 3
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. General-domain results. The source labels this column "↓(%)" but does not state its calculation.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy interdisciplinary question generation (Show measurement, test conditions, and source)
- Source value
- 79.8
- Score
- 79.8
- Metric
- Rel
- Unit
- %
- Benchmark version
- ACL 2024, Table 5
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 5 reports the interdisciplinary task; its ICL average uses a human-written sample in the prompt.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy interdisciplinary question generation (Show measurement, test conditions, and source)
- Source value
- 80
- Score
- 80
- Metric
- Rep
- Unit
- %
- Benchmark version
- ACL 2024, Table 5
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 5 reports the interdisciplinary task; its ICL average uses a human-written sample in the prompt.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy interdisciplinary question generation (Show measurement, test conditions, and source)
- Source value
- 79.9
- Score
- 79.9
- Metric
- Aver
- Unit
- %
- Benchmark version
- ACL 2024, Table 5
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. The paper evaluates generated questions for use in education. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion. Table 5 reports the interdisciplinary task; its ICL average uses a human-written sample in the prompt.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Dr.Academy interdisciplinary question generation (Show measurement, test conditions, and source)
- Source value
- 71.7
- Score
- 71.7
- Metric
- 0-shot
- Unit
- %
- Benchmark version
- ACL 2024, Table 5
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Dr.Academy paper (ACL Anthology, 2024)
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Interdisciplinary question generation, 0-shot setting. GPT-4 scores each question three times using binary 1/0 criteria; at least two scores of 1 meet a criterion.
- Context
- Tables 3 and 5 do not report evaluated N, confidence intervals, model-generation settings, or run date. The paper's separate manual evaluation sampled 1,000 questions per model and task; that is not the N for these automatic table rows.
Page 1 of 4
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 101 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
Falcon 180BLlama-3.2-90B-Vision-Instruct 26o3-pro 26Index-Homura-9B 25Trinity Nano Preview 25
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Falcon 180B, chosen by model line, developer, size, price and release date.
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about Falcon 180B
How large is the context window of Falcon 180B?
2,048 tokens. That is how much text Falcon 180B reads at once, about 0 A4 pages.
Page count as a rule of thumb from Artificial Analysis
GLM-5.1200K
DeepSeek-V3 0324164K
GLM-4.5131K
GPT OSS 120B131K
DeepSeek-V3.2128K
DeepSeek-R1128K
Falcon 180BRank 103 of 1032K
Can I run Falcon 180B locally?
Yes, the weights are open. With 180B parameters it needs a lot of memory, more a server than a laptop.
Stated by TII
- Open weights
- yes
- License
- Falcon License
- Parameters
- 180B
- active per token
- n/a
Sources and data date
Show all 10 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. |
| Output | Documented modalities. |
| Research status | Research date October 15, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | flattened-table-20 Note Falcon report Table 20's flattened/multi-column extraction does not unambiguously associate all numeric cells with model and task. Those ambiguous values are omitted rather than guessed. The paper also omits sample counts for these report tables. |
| Additional source | Hugging Face Falcon 180B evaluation retrospective (retrieved October 15, 2026) |
| Additional source | Falcon technical report (retrieved October 16, 2026) |
| Additional source | Dr.Academy paper (ACL Anthology, 2024) (retrieved October 19, 2026) |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |