Çöñţéẋţ éṽàļ
ķàþî šüþþļîéš ḿöđéļš ŵîţĥ þŕöĵéçţ çöñţéẋţ, îñçļüđîñĝ ŕéǫüîŕéđ ţéŕḿš, à ṽöîçé ĝüîđé àñđ îñšţŕüçţîöñš. Ţĥîš þàĝé ḿéàšüŕéš ĥöŵ ţĥàţ çöñţéẋţ àƒƒéçţš àđĥéŕéñçé ţö éàçĥ ŕéǫüîŕéḿéñţ. Îţ đöéš ñöţ àššéšš öṽéŕàļļ ţŕàñšļàţîöñ ǫüàļîţý. Ţĥé ƃàţçĥ éṽàļ ḿéàšüŕéš šţŕüçţüŕàļ îñţéĝŕîţý àñđ çöšţ.
Ĥöŵ îţ îš ḿéàšüŕéđ
À ţéšţ çöŕþüš îš ţŕàñšļàţéđ ţŵîçé ţĥŕöüĝĥ ţĥé þŕöđüçţîöñ þîþéļîñé: öñçé ƃàŕé (ŵîţĥöüţ çöñţéẋţ) àñđ öñçé šţééŕéđ (ŵîţĥ ţéŕḿš, à ṽöîçé þŕöƒîļé àñđ îñšţŕüçţîöñš). Ƃöţĥ þàššéš àŕé šçöŕéđ ŵîţĥ ķàþî’š đéţéŕḿîñîšţîç çĥéçķš: ţéŕḿ-çĥéçķ, đñţ-çĥéçķ, ṽöîçé-ṽöçàƃ-çĥéçķ àñđ þàţţéŕñ-çĥéçķ. Éàçĥ đîḿéñšîöñ ŕéþöŕţš àƃšöļüţé àđĥéŕéñçé (ţĥé þŕöþöŕţîöñ öƒ ŕéǫüîŕéḿéñţš šàţîéđ ŵîţĥ çöñţéẋţ) àñđ ļîƒţ (àđĥéŕéñçé ŵîţĥ çöñţéẋţ ḿîñüš àđĥéŕéñçé ŵîţĥöüţ îţ). Ĥîĝĥ àđĥéŕéñçé ŵîţĥ ñö ļîƒţ ḿéàñš ţĥé ḿöđéļ àļŕéàđý ḿéţ ţĥöšé ŕéǫüîŕéḿéñţš ŵîţĥöüţ ţĥé šüþþļîéđ çöñţéẋţ. Þöšîţîṽé ļîƒţ ḿéàšüŕéš ţĥé îḿþŕöṽéḿéñţ àţţŕîƃüţàƃļé ţö ţĥàţ çöñţéẋţ îñ ţĥîš çöŕþüš.
Ƒîẋţüŕéš ţéšţ ŕéǫüîŕéḿéñţš à ţŕàñšļàţîöñ ḿîĝĥţ öţĥéŕŵîšé ṽîöļàţé: àñ üñüšüàļ ŕéǫüîŕéđ ţéŕḿ, à þŕöđüçţ ñàḿé ţĥàţ ŕéšéḿƃļéš à çöḿḿöñ ñöüñ, à ƒöŕḿàļ ŕéĝîšţéŕ ƒöŕ çàšüàļ šöüŕçé ţéẋţ, öŕ à ƃàñ öñ éẋçļàḿàţîöñ ḿàŕķš. Đîšţŕàçţöŕš ţéšţ ŵĥéţĥéŕ ţĥé ḿöđéļ àþþļîéš ŕüļéš ţöö ƃŕöàđļý. Çöñƒļîçţîñĝ ŕüļéš ĥàṽé àñ éẋþļîçîţ þŕîöŕîţý; ƒöŕ éẋàḿþļé, à ŕéǫüîŕéđ çöḿþöüñđ çàñ ţàķé þŕéçéđéñçé öṽéŕ à ƃàñ öñ öñé öƒ îţš ŵöŕđš. Ŕéšüļţš àŕé ŕéþöŕţéđ šéþàŕàţéļý ƒöŕ ţéŕḿîñöļöĝý, ṽöîçé àñđ îñšţŕüçţîöñš šö šţŕéñĝţĥš îñ öñé đîḿéñšîöñ đö ñöţ öƃšçüŕé ŵéàķñéššéš îñ àñöţĥéŕ.
Šüƃĵéçţîṽé ṽöîçé ǫüàļîţîéš, îñçļüđîñĝ ŕéĝîšţéŕ, ñàţüŕàļñéšš àñđ ŕéšţŕàîñţ, àŕé šçöŕéđ ƃý àñ ĻĻḾ ĵüđĝé ƒŕöḿ à đéŕéñţ ḿöđéļ ƒàḿîļý. Ţĥé ĵüđĝé üšéš à ƒîẋéđ ýéš/ñö ŕüƃŕîç àñđ ŕéçéîṽéš ñö îñƒöŕḿàţîöñ àƃöüţ ţĥé ḿöđéļ öŕ ṽàŕîàñţ ţĥàţ þŕöđüçéđ ţĥé ţéẋţ. Šçöŕéš àŕé þüƃļîšĥéđ öñļý àƒţéŕ ḿéàšüŕéđ àĝŕééḿéñţ ŵîţĥ ĥüḿàñ ŕéṽîéŵéŕš éẋçééđš ţĥé šţàţéđ ţĥŕéšĥöļđ. Üñţîļ ţĥéñ, ţĥéý ŕéḿàîñ îñ ţĥé ŕüñ ŕéçöŕđ ƃüţ àŕé éẋçļüđéđ ƒŕöḿ ţĥîš þàĝé.
Ŵĥàţ ţĥé ḿéàšüŕéḿéñţ ƒöüñđ
Ţĥé ḿöđéļ ŵîţĥ ţĥé ĝŕéàţéšţ ḿéàšüŕéđ ļîƒţ îš haiku ( nb ), ŵîţĥ àñ öṽéŕàļļ îḿþŕöṽéḿéñţ öƒ +73.1pp.Éṽéŕý ḿéàšüŕéđ ḿöđéļ šĥöŵš þöšîţîṽé ļîƒţ öñ ţĥîš çöŕþüš.The weakest dimension with context applied is instruction at 97.6% steered adherence.
Ţàŕĝéţ de ḿéàšüŕéđ 2026-08-27 · 20 fixtures, 32 checks (17 terminology, 7 voice, 8 instruction) · çöŕþüš 66cc1693369f
haiku (claude-code) sonnet (claude-code) öþéñ đöţ = ŵîţĥöüţ çöñţéẋţ · ƒîļļéđ đöţ = ŵîţĥ çöñţéẋţ
| Ḿöđéļ | terminology ļîƒţ | voice ļîƒţ | instruction ļîƒţ | öṽéŕàļļ ļîƒţ | çöñţéẋţ ţöķ/þàšš | ļîƒţ / 1ķ çţẋ ţöķ | šţééŕéđ $/þàšš |
|---|
| haiku | +41.2pp | +42.9pp | +62.5pp | +46.9pp | — | — | — |
| sonnet | +47.1pp | +42.9pp | +75.0pp | +53.1pp | — | — | — |
Þéŕ-ţŕàþ ƃŕéàķđöŵñ
Àđĥéŕéñçé ŵîţĥ çöñţéẋţ, ƃý ţĥé ķîñđ öƒ ţŕàþ ţĥé ƒîẋţüŕé šéţ. À ḿöđéļ çàñ ĥöļđ ţĥé ţéŕḿš àñđ šţîļļ ţŕàñšļàţé ţĥé þŕöđüçţ ñàḿé; ţĥîš îš ŵĥéŕé ţĥàţ šĥöŵš.
haiku claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 67% | +66.7pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 100% | 100% | +0.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 13% | 100% | +87.5pp |
| terminology / term-conflict | 1 | 100% | 100% | +0.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / formality | 3 | 100% | 100% | +0.0pp |
| voice / vocab | 3 | 33% | 100% | +66.7pp |
| voice / vocab-conflict | 1 | 0% | 100% | +100.0pp |
sonnet claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 100% | 100% | +0.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 0% | 100% | +100.0pp |
| terminology / term-conflict | 1 | 100% | 100% | +0.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / formality | 3 | 100% | 100% | +0.0pp |
| voice / vocab | 3 | 33% | 100% | +66.7pp |
| voice / vocab-conflict | 1 | 0% | 100% | +100.0pp |
Ţàŕĝéţ en-GB ḿéàšüŕéđ 2026-07-17 · 16 fixtures, 20 checks (3 terminology, 5 voice, 12 instruction) · çöŕþüš 600fac47f9dc
eu.anthropic.claude-sonnet-4-6 (bedrock) haiku (claude-code) opus (claude-code) sonnet (claude-code) gemini-3.1-flash-lite (gemini) gemini-3.1-pro-preview (gemini) gemini-3.5-flash (gemini) öþéñ đöţ = ŵîţĥöüţ çöñţéẋţ · ƒîļļéđ đöţ = ŵîţĥ çöñţéẋţ
| Ḿöđéļ | terminology ļîƒţ | voice ļîƒţ | instruction ļîƒţ | öṽéŕàļļ ļîƒţ | çöñţéẋţ ţöķ/þàšš | ļîƒţ / 1ķ çţẋ ţöķ | šţééŕéđ $/þàšš |
|---|
| eu.anthropic.claude-sonnet-4-6 | +100.0pp | +100.0pp | +50.0pp | +70.0pp | 150 | +466.7pp | $0.014 |
| haiku | +100.0pp | +100.0pp | +41.7pp | +65.0pp | — | — | — |
| opus | +100.0pp | +100.0pp | +50.0pp | +70.0pp | — | — | — |
| sonnet | +100.0pp | +100.0pp | +50.0pp | +70.0pp | — | — | — |
| gemini-3.1-flash-lite | +100.0pp | +100.0pp | +50.0pp | +70.0pp | 137 | +510.9pp | $0.0011 |
| gemini-3.1-pro-preview | +100.0pp | +100.0pp | +50.0pp | +70.0pp | 137 | +510.9pp | $0.0076 |
| gemini-3.5-flash | +100.0pp | +100.0pp | +50.0pp | +70.0pp | 137 | +510.9pp | $0.0047 |
Þéŕ-ţŕàþ ƃŕéàķđöŵñ
Àđĥéŕéñçé ŵîţĥ çöñţéẋţ, ƃý ţĥé ķîñđ öƒ ţŕàþ ţĥé ƒîẋţüŕé šéţ. À ḿöđéļ çàñ ĥöļđ ţĥé ţéŕḿš àñđ šţîļļ ţŕàñšļàţé ţĥé þŕöđüçţ ñàḿé; ţĥîš îš ŵĥéŕé ţĥàţ šĥöŵš.
eu.anthropic.claude-sonnet-4-6 bedrock · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / spelling | 4 | 100% | 100% | +0.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / term | 3 | 0% | 100% | +100.0pp |
| voice / contractions | 4 | 0% | 100% | +100.0pp |
| voice / vocab | 1 | 0% | 100% | +100.0pp |
haiku claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 67% | +66.7pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / spelling | 4 | 100% | 100% | +0.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / term | 3 | 0% | 100% | +100.0pp |
| voice / contractions | 4 | 0% | 100% | +100.0pp |
| voice / vocab | 1 | 0% | 100% | +100.0pp |
opus claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / spelling | 4 | 100% | 100% | +0.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / term | 3 | 0% | 100% | +100.0pp |
| voice / contractions | 4 | 0% | 100% | +100.0pp |
| voice / vocab | 1 | 0% | 100% | +100.0pp |
sonnet claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / spelling | 4 | 100% | 100% | +0.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / term | 3 | 0% | 100% | +100.0pp |
| voice / contractions | 4 | 0% | 100% | +100.0pp |
| voice / vocab | 1 | 0% | 100% | +100.0pp |
gemini-3.1-flash-lite gemini · 2× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 6 | 0% | 100% | +100.0pp |
| instruction / exclamation | 6 | 0% | 100% | +100.0pp |
| instruction / spelling | 8 | 100% | 100% | +0.0pp |
| instruction / verbatim | 4 | 100% | 100% | +0.0pp |
| terminology / term | 6 | 0% | 100% | +100.0pp |
| voice / contractions | 8 | 0% | 100% | +100.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
gemini-3.1-pro-preview gemini · 2× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 6 | 0% | 100% | +100.0pp |
| instruction / exclamation | 6 | 0% | 100% | +100.0pp |
| instruction / spelling | 8 | 100% | 100% | +0.0pp |
| instruction / verbatim | 4 | 100% | 100% | +0.0pp |
| terminology / term | 6 | 0% | 100% | +100.0pp |
| voice / contractions | 8 | 0% | 100% | +100.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
gemini-3.5-flash gemini · 2× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 6 | 0% | 100% | +100.0pp |
| instruction / exclamation | 6 | 0% | 100% | +100.0pp |
| instruction / spelling | 8 | 100% | 100% | +0.0pp |
| instruction / verbatim | 4 | 100% | 100% | +0.0pp |
| terminology / term | 6 | 0% | 100% | +100.0pp |
| voice / contractions | 8 | 0% | 100% | +100.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
Ţàŕĝéţ fr ḿéàšüŕéđ 2026-08-27 · 19 fixtures, 31 checks (17 terminology, 6 voice, 8 instruction) · çöŕþüš 7d42625fd539
haiku (claude-code) sonnet (claude-code) öþéñ đöţ = ŵîţĥöüţ çöñţéẋţ · ƒîļļéđ đöţ = ŵîţĥ çöñţéẋţ
| Ḿöđéļ | terminology ļîƒţ | voice ļîƒţ | instruction ļîƒţ | öṽéŕàļļ ļîƒţ | çöñţéẋţ ţöķ/þàšš | ļîƒţ / 1ķ çţẋ ţöķ | šţééŕéđ $/þàšš |
|---|
| haiku | +52.9pp | +33.3pp | +75.0pp | +54.8pp | — | — | — |
| sonnet | +52.9pp | +16.7pp | +75.0pp | +51.6pp | — | — | — |
Þéŕ-ţŕàþ ƃŕéàķđöŵñ
Àđĥéŕéñçé ŵîţĥ çöñţéẋţ, ƃý ţĥé ķîñđ öƒ ţŕàþ ţĥé ƒîẋţüŕé šéţ. À ḿöđéļ çàñ ĥöļđ ţĥé ţéŕḿš àñđ šţîļļ ţŕàñšļàţé ţĥé þŕöđüçţ ñàḿé; ţĥîš îš ŵĥéŕé ţĥàţ šĥöŵš.
haiku claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 100% | 100% | +0.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 0% | 100% | +100.0pp |
| terminology / term-conflict | 1 | 0% | 100% | +100.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / formality | 3 | 100% | 100% | +0.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
| voice / vocab-conflict | 1 | 100% | 100% | +0.0pp |
sonnet claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 100% | 100% | +0.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 0% | 100% | +100.0pp |
| terminology / term-conflict | 1 | 0% | 100% | +100.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / formality | 3 | 100% | 100% | +0.0pp |
| voice / vocab | 2 | 50% | 100% | +50.0pp |
| voice / vocab-conflict | 1 | 100% | 100% | +0.0pp |
Ţàŕĝéţ nb ḿéàšüŕéđ 2026-08-27 · 17 fixtures, 26 checks (16 terminology, 2 voice, 8 instruction) · çöŕþüš daf7de0cd29f
haiku (claude-code) sonnet (claude-code) öþéñ đöţ = ŵîţĥöüţ çöñţéẋţ · ƒîļļéđ đöţ = ŵîţĥ çöñţéẋţ
| Ḿöđéļ | terminology ļîƒţ | voice ļîƒţ | instruction ļîƒţ | öṽéŕàļļ ļîƒţ | çöñţéẋţ ţöķ/þàšš | ļîƒţ / 1ķ çţẋ ţöķ | šţééŕéđ $/þàšš |
|---|
| haiku | +68.8pp | +100.0pp | +75.0pp | +73.1pp | — | — | — |
| sonnet | +50.0pp | +100.0pp | +50.0pp | +53.8pp | — | — | — |
Þéŕ-ţŕàþ ƃŕéàķđöŵñ
Àđĥéŕéñçé ŵîţĥ çöñţéẋţ, ƃý ţĥé ķîñđ öƒ ţŕàþ ţĥé ƒîẋţüŕé šéţ. À ḿöđéļ çàñ ĥöļđ ţĥé ţéŕḿš àñđ šţîļļ ţŕàñšļàţé ţĥé þŕöđüçţ ñàḿé; ţĥîš îš ŵĥéŕé ţĥàţ šĥöŵš.
haiku claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 100% | +100.0pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 50% | 100% | +50.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 0% | 100% | +100.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
sonnet claude-code · 1× þéŕ ṽàŕîàñţ
| đîḿéñšîöñ / ķîñđ | çĥéçķš | ƃàŕé | šţééŕéđ | ļîƒţ |
|---|
| instruction / digits | 3 | 0% | 33% | +33.3pp |
| instruction / exclamation | 3 | 0% | 100% | +100.0pp |
| instruction / verbatim | 2 | 100% | 100% | +0.0pp |
| terminology / dnt | 6 | 100% | 100% | +0.0pp |
| terminology / dnt-distractor | 1 | 100% | 100% | +0.0pp |
| terminology / term | 8 | 0% | 100% | +100.0pp |
| terminology / term-distractor | 1 | 100% | 100% | +0.0pp |
| voice / vocab | 2 | 0% | 100% | +100.0pp |
Ŕéþŕöđüçîñĝ ţĥîš
Ţĥé ĥàŕñéšš îš scripts/contexteval. À ŕüñ àĝàîñšţ ţĥé ƃüîļţ-îñ đéḿö šţüƃ éẋéŕçîšéš ţĥé ĥàŕñéšš àñđ ḿéàšüŕéš ñöţĥîñĝ àƃöüţ àñý ḿöđéļ; šüçĥ ŕüñš àŕé ḿàŕķéđ simulated àñđ àŕé éẋçļüđéđ ƒŕöḿ éṽéŕý çĥàŕţ öñ ţĥîš þàĝé. Àđĥéŕéñçé ṽàŕîéš ƃý ţàŕĝéţ ļàñĝüàĝé, šö ţĥé þüƃļîšĥéđ šŵééþ çöṽéŕš ḿöŕé ţĥàñ öñé.
make context-eval # demo stub: proves the harness, measures nothing
make context-eval-publish # the real sweep → this page's data