Part of Best Cheap LLM for High Volume

Best Open-Weight LLM for Self-Hosting

GLM 5.2 is the provisional first pick for Best Open-Weight LLM for Self-Hosting as of November 15, 2026, while independent task measurements are still limited.[1][2] It ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes. GLM 5.3 Flash is the next-ranked alternative. It ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.

About this recommendation

Updated
Evidence through
Sources
9
Revision
v139

Decision audit

Why this result

Inspect the inputs and the computed order behind the recommendation.

Models screened

24

live candidates

Evaluation feeds

4

task-weighted

Winner coverage

53%

intended feed weight

Largest provider share

4 of 8

Z.ai

Provisional source breadth. 6 citation families and 2 practitioner families provide attributed capability reports; 1 cautionary thread retained. The largest citation family contributes 57%.

Z.ai: GLM 5.2 leads the measured task comparison with limited independent task measurements. Community volume, release age and usage popularity add no ranking points. This category also weights its stated price or context criteria.

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.

winner: GLM 5.2
Evaluation feedWeightWinner resultField measured
price weight
67%
100/10024/24
LiveBench Instruction Following (2026-06-25)Observed 2026-10-18 · evaluation date not supplied · glm-5.2failing; last successful check 2026-10-18stale
0%
no current result0/24
Route reliabilitydisabled; last successful check unknownunavailable
0%
feed unavailable0/24
LMArena TextObserved 2026-11-11 · evaluation date not supplied · max effort
33%
#2624/24

Provider concentration

Each exact model is scored separately; provider identity is not a ranking input.

Z.ai50%
  • Z.ai4 models
  • deepseek2 models
  • Google1 model
  • OpenAI1 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.

  1. Z.ai · coverage 53% · provisional evidence

    #26 LMArena Text

  2. Z.ai · coverage 53% · provisional evidence

    #26 LMArena Text

  3. deepseek · coverage 53% · provisional evidence

    #26 LMArena Text

  4. Z.ai · coverage 53% · provisional evidence

    #20 LMArena Text

  5. Google · coverage 53% · provisional evidence

    #65 LMArena Text

  6. OpenAI · coverage 53% · provisional evidence

    #132 LMArena Text

  7. deepseek · coverage 53% · provisional evidence

    #67 LMArena Text

  8. Z.ai · coverage 53% · provisional evidence

    #115 LMArena Text

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.

  1. 01

    GLM 5.2Z.ai · $0.007 in / $6.80 out per M tokens · 1.0M context · open-weight

    GLM 5.2 ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.

    Best when: Its strongest result for this task is on LMArena Text, which measures general reasoning and instruction following.
    Source 1
    “Within the 147 catalog models matched to this evaluation: Ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.”
    LMArena text arenaOpen original ↗
  2. 02

    GLM 5.3 FlashZ.ai · $0.040 in / $0.078 out per M tokens · 1.0M context · open-weight

    GLM 5.3 Flash ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.

    Best when: Its strongest result for this task is on LMArena Text, which measures general reasoning and instruction following.
    Source 2
    “Within the 147 catalog models matched to this evaluation: Ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.”
    LMArena text arenaOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  3. 03

    DeepSeek V4.1 Flashdeepseek · $0.051 in / $0.20 out per M tokens · 1.0M context · open-weight

    DeepSeek V4.1 Flash ranks #26 on LMArena with Elo 1475 and scores 70.03% on LiveBench Instruction Following, with weights now published on Hugging Face after initial API-only preview.

    Best when: Strong instruction following: 70.03% on LiveBench places it at #25 of 59, ahead of many competitors on practical tasks like paraphrasing and summarization.
    Source 3
    “Within the 59 catalog models matched to this evaluation: Scores 70.03% on LiveBench Instruction Following (#25 of 59), including paraphrasing, simplifying, story generation, and summarization.”
    LiveBench Instruction FollowingOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  4. 04

    GLM 5.3Z.ai · $0.084 in / $3.60 out per M tokens · 1.0M context · open-weight

    GLM 5.3 ranks #20 of 147 on LMArena's overall text arena (Elo 1482), based on blind human preference votes.

    Best when: Its strongest result for this task is on LMArena Text, which measures general reasoning and instruction following.
    Source 4
    “Within the 147 catalog models matched to this evaluation: Ranks #20 of 147 on LMArena's overall text arena (Elo 1482), based on blind human preference votes.”
    LMArena text arenaOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  5. 05

    Gemma 4 26B A4B Google · $0.068 in / $0.22 out per M tokens · 262K context · open-weight

    Gemma 4 26B A4B ranks #65 on LMArena with Elo 1437, with community reports of successful CPU-only deployment on i5-8500 with 32GB RAM and MoE architecture providing value for memory-constrained use cases.

    Best when: Extreme hardware flexibility: reported running on i5-8500 with 32GB RAM and no GPU, making it accessible for minimal hardware setups.
    Source 5
    “Someone was able to run gemma-4-26B-A4B on an i5-8500 with 32 gb ram with NO GPU. Granted this is an extreme example these MoE models are value for money for a lot of use cases. https: www.reddit.com r LocalLLaMA s YontVNVRbL”

    More tips

    • Avoids token quota limitations: local deployment eliminates cloud service pauses and quota restrictions experienced with API-based workflows.
      Source 6
      “Not replaced but supplemented. For off-line coding current setup is pi + ds4-server + DeepSeek-V4-Flash REAP25 (on M2 Max 96gb). For simpler programming related (e.g. text2sql) as well as synthetic data generation, current best for me is llama.cpp + Gemma-4-26B-A4B (on gpu 7900xtx 24gb; sometimes nemotron-cascade-2-30b-a3b for 1M context). That and (dabbling now) auto-research uses lots of tokens. Used to get paused running out of token quotas all the time. The 1st local model I found somewhat…”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  6. 06

    gpt-oss-20bOpenAI · $0.030 in / $0.13 out per M tokens · 131K context · open-weight

    gpt-oss-20b ranks #132 of 147 on LMArena with Elo 1318, positioned as a free alternative to paid models for routine coding tasks though availability issues reported on some platforms.

    Best when: Free alternative for coding workflows: used as zero-cost substitute for paid models in Codex CLI + Ollama setups for Vela implementation work.
    Source 7
    “Goal Test whether the home desktop can use Codex CLI + Ollama + gpt-oss:20b for meaningful Vela implementation work, reserving paid Codex/ChatGPT for planning, difficult problems, and independent final review. This is a research/pilot phase, not a production agent rollout. Scope - Verify the local hardware/runtime configuration (AMD RX 7900 XTX, concurrent Vela dev services, Ollama, context window and memory/VRAM usage). Record supported Codex CLI / Ollama integration and versions…”
    aarondewindtOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  7. 07

    DeepSeek V4 Flash 0423deepseek · $0.090 in / $0.18 out per M tokens · 1.0M context · open-weight

    DeepSeek V4 Flash 0423 ranks #67 of 147 on LMArena's overall text arena (Elo 1436), based on blind human preference votes.

    Best when: Its strongest result for this task is on LMArena Text, which measures general reasoning and instruction following.
    Source 8
    “Within the 147 catalog models matched to this evaluation: Ranks #67 of 147 on LMArena's overall text arena (Elo 1436), based on blind human preference votes.”
    LMArena text arenaOpen original ↗

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  8. 08

    GLM 4.7 FlashZ.ai · $0.061 in / $0.40 out per M tokens · 200K context · open-weight

    GLM 4.7 Flash ranks #115 of 147 on LMArena's overall text arena (Elo 1365), based on blind human preference votes.

    Best when: Its strongest result for this task is on LMArena Text, which measures general reasoning and instruction following.
    Source 9
    “Within the 147 catalog models matched to this evaluation: Ranks #115 of 147 on LMArena's overall text arena (Elo 1365), based on blind human preference votes.”
    LMArena text arenaOpen original ↗

    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 Best Open-Weight LLM for Self-Hosting?
GLM 5.2 ranks first as of November 15, 2026. GLM 5.2 ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.[1]
What is the best alternative to GLM 5.2?
GLM 5.3 Flash ranks second. GLM 5.3 Flash ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.[2]
Can I self-host the top pick for Best Open-Weight LLM for Self-Hosting?
Yes. GLM 5.2 is open-weight, so it can run on your own hardware as well as through hosted routes.[1]

Sources

  1. 1

    “Within the 147 catalog models matched to this evaluation: Ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.”

    LMArena text arena · Benchmark · Nov 11, 2026
  2. 2

    “Within the 147 catalog models matched to this evaluation: Ranks #26 of 147 on LMArena's overall text arena (Elo 1475), based on blind human preference votes.”

    LMArena text arena · Benchmark · Nov 11, 2026
  3. 3

    “Within the 59 catalog models matched to this evaluation: Scores 70.03% on LiveBench Instruction Following (#25 of 59), including paraphrasing, simplifying, story generation, and summarization.”

    LiveBench Instruction Following · Benchmark · Oct 18, 2026
  4. 4

    “Within the 147 catalog models matched to this evaluation: Ranks #20 of 147 on LMArena's overall text arena (Elo 1482), based on blind human preference votes.”

    LMArena text arena · Benchmark · Nov 11, 2026
  5. 5

    “Someone was able to run gemma-4-26B-A4B on an i5-8500 with 32 gb ram with NO GPU. Granted this is an extreme example these MoE models are value for money for a lot of use cases. https: www.reddit.com r LocalLLaMA s YontVNVRbL”

    sbmthakur · Hacker News · Jun 16, 2026
  6. 6

    “Not replaced but supplemented. For off-line coding current setup is pi + ds4-server + DeepSeek-V4-Flash REAP25 (on M2 Max 96gb). For simpler programming related (e.g. text2sql) as well as synthetic data generation, current best for me is llama.cpp + Gemma-4-26B-A4B (on gpu 7900xtx 24gb; sometimes nemotron-cascade-2-30b-a3b for 1M context). That and (dabbling now) auto-research uses lots of tokens. Used to get paused running out of token quotas all the time. The 1st local model I found somewhat…”

    ljosifov · Hacker News · Jun 16, 2026
  7. 7

    “Goal Test whether the home desktop can use Codex CLI + Ollama + gpt-oss:20b for meaningful Vela implementation work, reserving paid Codex/ChatGPT for planning, difficult problems, and independent final review. This is a research/pilot phase, not a production agent rollout. Scope - Verify the local hardware/runtime configuration (AMD RX 7900 XTX, concurrent Vela dev services, Ollama, context window and memory/VRAM usage). Record supported Codex CLI / Ollama integration and versions…”

    aarondewindt · GitHub · Nov 4, 2026
  8. 8

    “Within the 147 catalog models matched to this evaluation: Ranks #67 of 147 on LMArena's overall text arena (Elo 1436), based on blind human preference votes.”

    LMArena text arena · Benchmark · Nov 11, 2026
  9. 9

    “Within the 147 catalog models matched to this evaluation: Ranks #115 of 147 on LMArena's overall text arena (Elo 1365), based on blind human preference votes.”

    LMArena text arena · Benchmark · Nov 11, 2026

Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.