Model decision table
Rankings with the tradeoffs left in.
Scan the full catalog, narrow it to your workload, and compare up to three models. Specifications are source-reviewed; fit scores are editorial and their weighting is explicit.
Full catalog
Compare models
Showing 19 models
| Compare | # | Model | Editorial fit | Coding | Reasoning | Tool use | Context | Price / 1M tokens | Evidence |
|---|---|---|---|---|---|---|---|---|---|
| 01 | Claude Fable 5Anthropic · AvailableLong-running agents · Highest-capability tasks | 9.9editorial | 9.9 | 9.9 | 9.8 | 1M | $10→$50 | ||
| 02 | GPT-5.5OpenAI · AvailableComplex reasoning · Coding | 9.8editorial | 9.8 | 9.8 | 9.7 | 1M | $5→$30 | ||
| 03 | Claude Opus 4.8Anthropic · AvailableComplex reasoning · Agentic coding | 9.8editorial | 9.8 | 9.8 | 9.6 | 1M | $5→$25 | ||
| 04 | GPT-5.4OpenAI · AvailableCoding · Agents | 9.7editorial | 9.8 | 9.5 | 9.7 | 1M | $2.5→$15 | ||
| 05 | Claude Sonnet 5Anthropic · AvailableBalanced performance · Production workloads | 9.5editorial | 9.6 | 9.5 | 9.3 | 1M | $3→$15Tier caveat | ||
| 06 | GPT-5.2-CodexOpenAI · AvailableCoding-focused tasks · Type inference | 9.5editorial | 9.7 | 9.3 | 9.4 | 400K | $1.75→$14 | ||
| 07 | Gemini 3.1 Pro PreviewGoogle · PreviewMultimodal tasks · Long context | 9.4editorial | 9.5 | 9.5 | 9.3 | 1M | $2→$12Tier caveat | ||
| 08 | Gemini 3.5 FlashGoogle · AvailableFast multimodal agents · Search grounding | 9.4editorial | 9.4 | 9.4 | 9.4 | 1M | $1.5→$9Tier caveat | ||
| 09 | Kimi K2.5Moonshot AI · AvailableVisual coding · Long context | 9.3editorial | 9.4 | 9.3 | 9.2 | 256K | $0.6→$2.5 | ||
| 10 | GPT-5.2OpenAI · AvailableGeneral-purpose · Balanced tasks | 9.2editorial | 9.3 | 9.2 | 9.0 | 400K | $1.75→$14 | ||
| 11 | GPT-OSS-120BOpenAI · AvailableSelf-hosted · Privacy | 9.2editorial | 9.3 | 9.2 | 9.0 | 128K | $0→$0Tier caveat | ||
| 12 | GLM-5Zhipu AI · AvailableBilingual (CN/EN) · Value-focused | 9.2editorial | 9.2 | 9.3 | 9.0 | 205K | $0.5→$2 | ||
| 13 | Grok 4.3xAI · AvailableLong context · Web search | 9.2editorial | 9.2 | 9.3 | 9.0 | 1M | $1.25→$2.5 | ||
| 14 | Qwen3 Max 2026-01-23Alibaba · AvailableMultilingual · Enterprise | 9.1editorial | 9.2 | 9.2 | 8.9 | 262K | $1.2→$6 | ||
| 15 | DeepSeek V4 ProDeepSeek · AvailableBudget coding · High-volume | 9.1editorial | 9.1 | 9.2 | 8.9 | 1M | $0.435→$0.87Tier caveat | ||
| 16 | MiniMax M2.5MiniMax · AvailableReal-world productivity · Cost-sensitive | 9.1editorial | 9.1 | 9.2 | 8.9 | 196K | $0.3→$1.2 | ||
| 17 | Mistral Medium 3.5Mistral · AvailableEuropean compliance · Agentic coding | 9.0editorial | 9.1 | 9.1 | 8.9 | 128K | $2→$6 | ||
| 18 | Llama 4 MaverickMeta · AvailableOpen weights · Multimodal | 9.0editorial | 9.0 | 9.1 | 8.7 | 1M | $2→$8Tier caveat | ||
| 19 | Claude Haiku 4.5Anthropic · AvailableFast responses · High-volume tasks | 8.8editorial | 8.9 | 8.8 | 8.7 | 200K | $1→$5 |
No models match these filters.
Reset the filters or broaden the workload.
Score methodology
Orientation, not manufactured certainty.
The editorial fit score weights coding at 40%, reasoning at 35%, and tool use at 25%. It is a compact decision aid built from the dimensions tracked in this catalog—not a claim that one model is universally best.
Prices show input then output cost per one million tokens. Provider tiers, caching, batches, and hosts can change the actual bill. Open the evidence links and caveat labels before committing to a production choice.
Inspect the source log and review policy →