Part of Best LLM for Vision & Documents
Best LLM for OCR & Documents
Gemini 3.8 Flash is the provisional first pick for OCR & Documents as of October 20, 2026, while independent task measurements are still limited.[1][2][3][4] Ranks #2 on LMArena Vision's OCR category with Elo 1316, though production use shows duplicate and contradictory extraction when chunking multi-page documents. Watch out: Watch for duplicate entries and conflicting field values when processing multi-page PDFs with mixed OCR strategies, as seen with 18 duplicate product entries and contradictory brand/specification data in a 5-page retail catalog. Gemini 3.7 Flash is the next-ranked alternative. Ranks #5 on LMArena Vision's OCR category with Elo 1308 and #4 on the broader vision arena, showing strong general image understanding alongside competitive OCR performance.
- Best open-weight pick: GLM 5.3 Flash (#7)
- Compare Gemini 3.8 Flash vs Gemini 3.7 Flash →
About this recommendation
- Updated
- Evidence through
- Sources
- 14
- Revision
- v104
Decision audit
Why this result
Inspect the inputs and the computed order behind the recommendation.
Models screened
24
live candidates
Evaluation feeds
5
task-weighted
Winner coverage
45%
intended feed weight
Largest provider share
2 of 8
Anthropic
Google: Gemini 3.8 Flash leads the measured task comparison with limited independent task measurements. Community volume, release age and usage popularity add no ranking points.
Real-world track record is still developing. The recommendation can lead on independent measurements while practitioner reports accumulate.
Sources evaluated
These are the weights used in scoring. Missing measurements are omitted from each model’s average. A feed with no result for any model released in the last 60 days is stale and does not score until it catches up.
| Evaluation feed | Weight | Winner result | Field measured |
|---|---|---|---|
| LMArena Vision OCRObserved 2026-10-19 · evaluation date not supplied · high effort | 46% | #2 | 24/24 |
| LMArena Document | 31% | not measured | 8/24 |
| VLMEvalKit tasksdisabled; last successful check unknownunavailable | 0% | feed unavailable | 0/24 |
| LMArena VisionObserved 2026-10-19 · evaluation date not supplied · high effort | 23% | #8 | 24/24 |
| Structured-output evaldisabled; last successful check unknownunavailable | 0% | feed unavailable | 0/24 |
Provider concentration
Each exact model is scored separately; provider identity is not a ranking input.
- Anthropic2 models
- Google2 models
- Meta2 models
- OpenAI1 model
- Z.ai1 model
Decision table
Every published model is shown in computed order. Practitioner sources are distinct community threads, not the citations repeated in the prose below.
Google · coverage 45% · provisional evidence
#2 LMArena Vision OCR · #8 LMArena Vision
Google · coverage 45% · provisional evidence
#4 LMArena Vision · #5 LMArena Vision OCR
Meta · coverage 45% · provisional evidence
#6 LMArena Vision · #6 LMArena Vision OCR
Anthropic · coverage 45% · provisional evidence
#6 LMArena Vision · #8 LMArena Vision OCR
Anthropic · coverage 65% · provisional evidence
#2 LMArena Document · #10 LMArena Vision
OpenAI · coverage 45% · provisional evidence
#5 LMArena Vision · #15 LMArena Vision OCR
Z.ai · coverage 45% · provisional evidence
#24 LMArena Vision · #25 LMArena Vision OCR
Meta · coverage 65% · provisional evidence
#6 LMArena Vision OCR · #9 LMArena Vision
Task score averages the available, task-weighted measurements. Benchmark positions compare models matched to our catalog. Models measured by at least two independent evaluator families are considered first. Missing tests, practitioner report counts, popularity and release age add no points. Coverage describes evidence breadth, not a probability of correctness. Scores within two points are shown as close under this policy; this is not a statistical confidence interval. When the leading scores are that close, the model measured by more evaluator families ranks first.
- 01
Gemini 3.8 FlashGoogle · $0.83 in / $4.13 out per M tokens · 1.0M context
Ranks #2 on LMArena Vision's OCR category with Elo 1316, though production use shows duplicate and contradictory extraction when chunking multi-page documents.
Best when: Use for single-image OCR where raw text extraction accuracy matters most, as it scores second-highest in blind human preference for reading text from images.Watch out for
- Watch for duplicate entries and conflicting field values when processing multi-page PDFs with mixed OCR strategies, as seen with 18 duplicate product entries and contradictory brand/specification data in a 5-page retail catalog.
- 02
Gemini 3.7 FlashGoogle · $0.75 in / $3.75 out per M tokens · 1.0M context
Ranks #5 on LMArena Vision's OCR category with Elo 1308 and #4 on the broader vision arena, showing strong general image understanding alongside competitive OCR performance.
Best when: Deploy for OCR tasks requiring balanced performance across general vision and text extraction, given its top-5 standing in both categories.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 03
Muse Spark 1.2Meta · $1.25 in / $4.25 out per M tokens · 1.0M context
Ties for #6 on LMArena Vision's OCR category with Elo 1307, with matching rank in the broader vision arena.
Best when: Consider when you need consistent performance across vision and OCR without specialization in either, as it holds identical ranks in both evaluation categories.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 04
Claude Opus 5.5Anthropic · $4 in / $20 out per M tokens · 1M context
Ranks #8 on LMArena Vision's OCR category with Elo 1306, trailing its vision arena rank by two positions.
Best when: Use when OCR is one component of a broader vision pipeline, as its general image understanding slightly outperforms its text extraction ranking.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 05
Claude Fable 5.1Anthropic · $10 in / $50 out per M tokens · 1M context
Ranks #2 on LMArena's document arena with score 1513, the strongest document-specific result in the pool, despite mid-tier OCR and vision rankings.
Best when: Prioritize for end-to-end document processing workflows where layout understanding and document structure matter more than raw OCR speed, given its #2 document arena standing.Watch out for
- Expect lower blind preference scores on pure OCR tasks compared to document tasks, as it ranks #10 in OCR category versus #2 in document arena.
- 06
GPT-6.1 SolOpenAI · $2 in / $10 out per M tokens · 1.1M context
Ranks #5 on LMArena's vision arena with Elo 1294 but drops to #15 on OCR-specific evaluation, showing stronger general vision than text extraction preference.
Best when: Its strongest result for this task is on LMArena Vision, which measures reading images, screenshots, and documents.Watch out for
- Avoid when pure text extraction accuracy is the primary criterion, as its OCR category rank of #15 with Elo 1297 falls below top-tier alternatives.
- 07
GLM 5.3 FlashZ.ai · $0.047 in / $0.16 out per M tokens · 1.0M context · open-weight
GLM 5.3 Flash ranks #24 of 81 on LMArena's vision arena (Elo 1271), based on human preference on image-understanding tasks.
Best when: Its strongest result for this task is on LMArena Vision, which measures reading images, screenshots, and documents.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 08
Muse Spark 1.3Meta · $1.25 in / $4.25 out per M tokens · 1.0M context
Muse Spark 1.3 ranks #6 of 75 on LMArena Vision's OCR category (Elo 1307), based on blind human preference for reading text out of images.
Best when: Its strongest result for this task is on LMArena Vision OCR, which measures reading text out of images and scans.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
Frequently asked
- What is the best model for OCR & Documents?
- Gemini 3.8 Flash ranks first as of October 20, 2026. Ranks #2 on LMArena Vision's OCR category with Elo 1316, though production use shows duplicate and contradictory extraction when chunking multi-page documents.[1][2]
- What should I watch out for with Gemini 3.8 Flash?
- Watch for duplicate entries and conflicting field values when processing multi-page PDFs with mixed OCR strategies, as seen with 18 duplicate product entries and contradictory brand/specification data in a 5-page retail catalog.[2]
- What is the best alternative to Gemini 3.8 Flash?
- Gemini 3.7 Flash ranks second. Ranks #5 on LMArena Vision's OCR category with Elo 1308 and #4 on the broader vision arena, showing strong general image understanding alongside competitive OCR performance.[3][4]
- Which open-weight model ranks highest for OCR & Documents?
- GLM 5.3 Flash is the highest-ranked open-weight model, at #7. GLM 5.3 Flash ranks #24 of 81 on LMArena's vision arena (Elo 1271), based on human preference on image-understanding tasks.[5]
Sources
- 1
“Within the 75 catalog models matched to this evaluation: Ranks #2 of 75 on LMArena Vision's OCR category (Elo 1316), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 2
fix(knowledge): 混合 OCR 產生重複頁面與矛盾商品條目（5 頁 PDF 出現 16 組頁面標題、92/74 商品重複）
Original-language excerpt:“現象 評測租戶知識庫「DM」的 DM_家樂福_采集漫旅_商品頁_p11-15.pdf（5 頁）以 Gemini 3.8 Flash 走混合 OCR（切片 2x3 + 整頁）處理後，落庫內容出現： - 16 組 【頁面標題】（其中 10 組為「不詳」），但原始 PDF 只有 5 頁 - 92 筆 商品： 條目，只有 74 筆唯一 — 18 筆重複 - 重複條目的欄位互相矛盾，例如同一商品「鮮剖100%椰子汁」出現兩次： - 一處 品牌：半天水、規格：含椰肉/純天然 520毫升×3、原價：120~129 - 另一處 品牌：鮮剖、規格：520毫升×3、原價：120-129元、備註：含椰肉/純天然 - 前者所在區段有完整頁面標題「單筆購買第12~15頁…」，後者所在區段頁面標題為「不詳」 已排除 chunk overlap 造成的假重複：把 chunk 依 chunk_index 串接後長度等於各 chunk 長度總和（無重疊），上述統計是文件本體的真實內容。 也已排除 worker 跑舊版：VM `poc-rag-v…”
larry610881 · GitHub · Sep 8, 2026 - 3
“Within the 75 catalog models matched to this evaluation: Ranks #5 of 75 on LMArena Vision's OCR category (Elo 1308), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 4
“Within the 81 catalog models matched to this evaluation: Ranks #4 of 81 on LMArena's vision arena (Elo 1295), based on human preference on image-understanding tasks.”
LMArena vision arena · Benchmark · Oct 19, 2026 - 5
“Within the 81 catalog models matched to this evaluation: Ranks #24 of 81 on LMArena's vision arena (Elo 1271), based on human preference on image-understanding tasks.”
LMArena vision arena · Benchmark · Oct 19, 2026 - 6
“Within the 81 catalog models matched to this evaluation: Ranks #6 of 81 on LMArena's vision arena (Elo 1293), based on human preference on image-understanding tasks.”
LMArena vision arena · Benchmark · Oct 19, 2026 - 7
“Within the 75 catalog models matched to this evaluation: Ranks #6 of 75 on LMArena Vision's OCR category (Elo 1307), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 8
“Within the 81 catalog models matched to this evaluation: Ranks #6 of 81 on LMArena's vision arena (Elo 1293), based on human preference on image-understanding tasks.”
LMArena vision arena · Benchmark · Oct 19, 2026 - 9
“Within the 75 catalog models matched to this evaluation: Ranks #8 of 75 on LMArena Vision's OCR category (Elo 1306), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 10
“Within the 34 catalog models matched to this evaluation: Ranks #2 of 34 on LMArena's document arena (score 1513), based on blind preference for document tasks.”
LMArena document arena · Benchmark · Sep 13, 2026 - 11
“Within the 75 catalog models matched to this evaluation: Ranks #10 of 75 on LMArena Vision's OCR category (Elo 1303), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 12
“Within the 81 catalog models matched to this evaluation: Ranks #5 of 81 on LMArena's vision arena (Elo 1294), based on human preference on image-understanding tasks.”
LMArena vision arena · Benchmark · Oct 19, 2026 - 13
“Within the 75 catalog models matched to this evaluation: Ranks #15 of 75 on LMArena Vision's OCR category (Elo 1297), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026 - 14
“Within the 75 catalog models matched to this evaluation: Ranks #6 of 75 on LMArena Vision's OCR category (Elo 1307), based on blind human preference for reading text out of images.”
LMArena Vision OCR · Benchmark · Oct 19, 2026
Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.