Undominated.ai

Model comparison · accepted catalogue facts 2026-10-21

gpt-oss-120b vs gpt-oss-20b

Neither model has both an equal-or-higher measured score and an equal-or-lower estimated price in this pair. Retaining your current model is a valid outcome.

This comparison was published earlier and keeps its address. These models are no longer adjacent on the default capability-price frontier. This is a direct pair comparison; global frontier membership is not its verdict.

gpt-oss-120b

LMArena general · Elo · higher is better

Scale starts at 1040 Elo

1365.4

Balanced effective USD / 1M tokens · lower is better

$0.060/M

gpt-oss-20b

LMArena general · Elo · higher is better

Scale starts at 1040 Elo

1287.4

Balanced effective USD / 1M tokens · lower is better

$0.036/M

In-Depth Analysis

gpt-oss-120b vs gpt-oss-20b: Measured Technical Breakdown

Comprehensive comparison across measured human-preference evaluation, effective production token costs, architectural limits, and migration trade-offs.

Measured Capability (LMArena)
gpt-oss-120b leads by 78.0 Elo

gpt-oss-120b: 1365.4 ±4.35 · gpt-oss-20b: 1287.4 ±6.39 Statistically significant preference separation

Cost Comparison Across Canonical Production Workloads

Rates represent effective cost per 1M tokens under standard token blends. Cached prompt reads are factored according to workload context reuse.

Workload & Ratiogpt-oss-120bgpt-oss-20bCost DifferenceAdvantage
Balanced 3 tokens in per 1 out$0.060 /1M tokens$0.036 /1M tokens$0.024 (40% diff)gpt-oss-20b costs less
Summarise Huge input, tiny output$0.032 /1M tokens$0.019 /1M tokens$0.013 (40% diff)gpt-oss-20b costs less
Chat Long cached system prompt$0.074 /1M tokens$0.044 /1M tokens$0.029 (40% diff)gpt-oss-20b costs less
Code gen Output dominates$0.101 /1M tokens$0.060 /1M tokens$0.040 (40% diff)gpt-oss-20b costs less
Agentic Context replayed each turn$0.039 /1M tokens$0.023 /1M tokens$0.016 (40% diff)gpt-oss-20b costs less

Architectural & Operational Specifications

Specificationgpt-oss-120bgpt-oss-20bOperational Note
Context window131,072 tokens131,072 tokens—
Max output generation65,536 tokens32,768 tokensgpt-oss-120b allows longer single-turn completions
Base input price$0.030 / 1M tokens$0.018 / 1M tokensgpt-oss-20b input rate is cheaper
Base output price$0.150 / 1M tokens$0.090 / 1M tokensgpt-oss-20b output rate is cheaper
Cached input price$0.015 / 1M tokens$0.0090 / 1M tokensgpt-oss-20b offers lower cached prompt cost
Reasoning & thinking tokensSupported; standard ratesSupported; standard rates—
Input modalitiestexttext—
Tool calling / function useSupportedSupported—
Open weights & licenceOpen weights (apache-2.0)Proprietary (API-only)gpt-oss-120b allows independent hosting or local deployment
Provider & reseller marketOpenAI · 24 provider offers (11.67× spread)OpenAI · 11 provider offers (4.17× spread)—

When to Choose gpt-oss-120b

  • Higher measured general capability (1365.4 Elo on human evaluation).
  • Higher maximum completion limit (65,536 tokens).
  • Open weights availability allows private self-hosting and on-premise execution.
  • Direct integration with OpenAI tooling and API features.

When to Choose gpt-oss-20b

  • Direct integration with OpenAI tooling and API features.
Risks when moving from gpt-oss-120b → gpt-oss-20b:
  • Output limit: 65,536 → 32,768 tokens.

Detailed Evidence Passports & Independent Audits

gpt-oss-120b Full model card, pricing ladder & provider offers → gpt-oss-20b Full model card, pricing ladder & provider offers → Alternatives to gpt-oss-120b Better-and-cheaper dominance audit → Alternatives to gpt-oss-20b Better-and-cheaper dominance audit →

Compare · model decision

Choose two or three models

Inspect the differences, estimate your workload and keep the evidence. All standard catalogue models are available, including unrated ones.

Catalogue facts
2026-10-21

2 / 3 selected · 0 requirements
Open public comparison link
Public workload and requirements

These settings are included in the share link. The fixed-request scenario and private evaluation data below are excluded.

A ↔ B

Neither model has both an equal-or-higher measured score and an equal-or-lower estimated price in this pair. Retaining your current model is a valid outcome.

Estimated blend: 75% input, 25% output; 0% cached input. Base context tier. Prompt length selects the billing tier. An unlisted cache-read rate uses that tier’s full input rate in this blend; missing required input or output rates prevent an estimate. The fixed-request estimate below validates actual context and output limits and requires a published cache-read rate when cache hits are assumed.

What changes between these models

Quality first, then estimated price. LMArena scores are human-preference measurements, not a benchmark of your application. A missing score is unrated.

  • gpt-oss-120b · Meets the stated model requirements
  • gpt-oss-20b · Meets the stated model requirements
Material differences and unknown evidence. Deltas compare each model with model A. Unknown facts remain visible.
Measured factA gpt-oss-120bB gpt-oss-20b
LMArena general score (Elo)1,365.41,287.4Δ vs A: -78 Elo
Score interval half-width (± Elo)4.356.39Δ vs A: +2.04 Elo
LMArena document score (Elo)UnknownUnknown
Estimated effective USD / 1M tokens$0.06$0.036Δ vs A: -0.024 USD/M
Base input USD / 1M tokens$0.03$0.018Δ vs A: -0.012 USD/M
Base output USD / 1M tokens$0.15$0.09Δ vs A: -0.06 USD/M
Base cached-input USD / 1M tokens$0.015$0.009Δ vs A: -0.006 USD/M
Output limit (tokens)65,53632,768Δ vs A: -32,768 tokens
Open weightsSupportedUnknown
Recorded licenceapache-2.0Unknown
Distinct sellers in the catalogue2210Δ vs A: -12
Recorded retirement dateUnknownUnknown
8 shared measured facts
Shared measured facts. Deltas compare each model with model A. Unknown facts remain visible.
Measured factA gpt-oss-120bB gpt-oss-20b
Benchmark reasoning effortdefaultdefault
Context window (tokens)131,072131,072Δ vs A: 0 tokens
Input modalitiestexttext
Tool useSupportedSupportedΔ vs A: 0
ReasoningSupportedSupportedΔ vs A: 0
Model providerOpenAIOpenAI
Serving precision differs across offersSupportedSupportedΔ vs A: 0
Retirement evidenceNone announced; not a guaranteeNone announced; not a guarantee

Provider counts and precision flags describe catalogue offers. They do not establish the precision, availability, region, tool behaviour or limits of your chosen endpoint. Open each model’s provider table before switching.

Complete price ladders and billing conditions

gpt-oss-120b

Rates below are USD per 1M tokens. A threshold selects that whole input/output rate pair.

Input-token thresholdInputOutputCached input
Base, up to and including first threshold$0.03$0.15$0.015

gpt-oss-20b

Rates below are USD per 1M tokens. A threshold selects that whole input/output rate pair.

Input-token thresholdInputOutputCached input
Base, up to and including first threshold$0.018$0.09$0.009
Migration preflight · compare each candidate with model A

Keeping gpt-oss-120b is a valid decision. No pairwise comparison certifies a drop-in replacement.

gpt-oss-120b → gpt-oss-20b

Known loss or failed requirement

  • Output limit: 65,536 → 32,768 tokens.
  • Provider endpoint, serving precision, availability and task success need your own verification. Catalogue model capabilities do not certify an endpoint.
Check alternatives to gpt-oss-120b

Estimate a fixed request shape

Private to this tab unless you save locally or explicitly export private details. These are estimates, not observed invoices. Cross-model token counts and task success are assumptions.

0.1 extra attempts means 10 extra full attempts per 100 requests. Cache writes, tool charges, taxes and unlisted charges are excluded.

  • gpt-oss-120bEstimated total: $0.135 1,000 full attempts · base rates
  • gpt-oss-20bEstimated total: $0.081 1,000 full attempts · base rates
Prompt and output sensitivity · sampled scenarios

Every point recalculates the entire request shape. Tier-boundary samples include one token below, at and above each threshold. Ordering between samples is not guaranteed. Lowest known cost excludes unknown estimates and is not a replacement recommendation.

Vary input; hold output, requests, cache and retries fixed.
Input tokensgpt-oss-120bgpt-oss-20bCoverage
2,000$0.135$0.081Lowest knownAll computed
4,096$0.19788$0.118728Lowest knownAll computed
32,768$1.05804$0.634824Lowest knownAll computed
131,072ineligibleInput plus output exceeds the published context window.ineligibleInput plus output exceeds the published context window.Incomplete
Vary output; hold input, requests, cache and retries fixed.
Output tokensgpt-oss-120bgpt-oss-20b
128$0.0792$0.04752
2,048$0.3672$0.22032
8,192$1.2888$0.77328

Bring your observed evaluation results

JSON is parsed locally. Prompts, filenames, individual responses and extra fields are not retained. Import only aggregate measurements for selected models. Task/version labels do not establish that two evaluations used the same dataset.

Expected format
[{"model":"provider/exact-model-id","task":"support-answer","version":"rubric-1","attempts":100,"successes":84,"costUsd":2.4,"latencyMs":950}]

Use the exact catalogue ID. Replace the example values with observed measurements. Omit costUsd or latencyMs when they are unknown. No public score is inferred from these observations.

Keep the decision and its evidence

A portable record retains selected facts, complete tier ladders, public requirements and the publication manifest. It also embeds the full permitted public catalogue to verify the input artifact checksum. Files are larger for this reason (up to 2 MB); local storage is limited to 12 records. Team review is local to this browser; there is no hosted collaboration.

Saving is opt-in and includes private review fields. Shared comparison links never include them. Version 1 Check references remain separate and are not overwritten or guessed into this format.

Saved local decisions · 0

No version 2 decision records are stored in this browser.

Questions this page answers

Is gpt-oss-120b a better-and-cheaper replacement for gpt-oss-20b?

Neither model has both an equal-or-higher measured score and an equal-or-lower estimated price in this pair. Retaining your current model is a valid outcome.

Which model is higher rated: gpt-oss-120b or gpt-oss-20b?

gpt-oss-120b scores higher on LMArena general (1365.4 ±4.4) compared to gpt-oss-20b (1287.4 ±6.4), a margin of 78.0 Elo points. The measured confidence intervals do not overlap, demonstrating a statistically significant preference lead on human evaluations.

Which model is more cost-effective: gpt-oss-120b or gpt-oss-20b?

On a standard balanced workload (3 prompt tokens per 1 generated token), gpt-oss-20b costs $0.036 per 1M tokens versus $0.060 for gpt-oss-120b. gpt-oss-120b is approximately 1.7× the price of gpt-oss-20b. Base input rates are $0.030 for gpt-oss-120b and $0.018 for gpt-oss-20b; base output rates are $0.150 for gpt-oss-120b and $0.090 for gpt-oss-20b.

How do their context windows and generation limits compare?

Both gpt-oss-120b and gpt-oss-20b provide an identical 131,072-token context window. Maximum output generation limits are 65,536 tokens for gpt-oss-120b and 32,768 tokens for gpt-oss-20b.

What capabilities are lost when switching between gpt-oss-120b and gpt-oss-20b?

Switching from gpt-oss-120b to gpt-oss-20b involves: Output limit: 65,536 → 32,768 tokens. Run test suites on your specific prompt templates and tool declarations before cutting over production traffic.

Can gpt-oss-120b or gpt-oss-20b be self-hosted or deployed privately?

gpt-oss-120b provides open weights, enabling self-hosting on private GPU infrastructure, local quantization, and private data pipelines. Managed inference is also offered across third-party API providers (24 for gpt-oss-120b, 11 for gpt-oss-20b).

What is included in this comparison?

Accepted model capabilities, context and output limits, complete price-tier ladders, independent LMArena measurements and missing evidence. Provider endpoint precision and application success still need verification.

Can I compare a different workload or keep a decision record?

Yes. Change the public workload, prompt length and requirements in the comparison workspace. Fixed-request scenarios, aggregate evaluations and review notes stay local unless explicitly exported.

Evidence & Ask
Markdown for LLMs